PWP-084 Vertical Integration in Gasoline Supply: an Empirical Test

Home , Double marginalization, Upstream price

PWP-084

Vertical Integration in Gasoline Supply: An Empirical Test of Raising Rivals' Costs

Richard Gilbert and Justine Hastings

July, 2001

This paper is part of the working papers series of the Program on Workable Energy Regulation (POWER). POWER is a program of the University of California Energy Institute, a multicampus research unit of the University of California, located on the Berkeley campus. University of California Energy Institute 2539 Channing Way Berkeley, California 94720-5180 www.ucei.org Vertical Integration in Gasoline Supply: An Empirical Test of Raising Rivals’ Costs

Richard Gilbert and Justine Hastings1 June 2001

Abstract

This paper explores the relationship between the structure of the market for the reÞning and distribution of gasoline and the wholesale price of unbranded gasoline sold to independent gasoline retailers. Theoretically, the eﬀectofanincreaseinverticalinte- gration is ambiguous because opposing forces act to increase and decrease wholesale prices. We empirically examine the eﬀects of vertical and horizontal market structures on wholesale prices using both a broad panel and an event analysis. The panel covers twenty-six metropolitan areas from January 1993 through June 1997. The event is a merger of Tosco and Unocal in 1997 that changed the vertical and horizontal structure of thirteen West Coast metropolitan areas. Both data sets show that an increase in the degree of vertical integration is associated with higher wholesale prices.

1University of California at Berkeley. We are grateful to John DiNardo, Michael Katz, Paul Ruud, Michael Whinston, and participants in the Industrial Organization Workshop at the University of California at Berkeley for helpful comments. This research was supported by a grant from the University of California Energy Institute.

1 1Introduction

Vertical integration can promote efficient trade and investment by mitigating the contractual hazards that can occur in markets characterized by speciÞc assets and bargaining among Þrms with market power (Williamson, 1975). However, vertical integration also can create incentives for Þrms to impose costs on rivals. This is a major thrust of the “post-Chicago” school of competition policy, which emphasizes the potential competitive hazards associated with vertical arrangements. For example, Ordover, Salop, and Saloner (1990) (henceforth OSS) show that a vertically integrated Þrm has an incentive to charge more for an intermediate good than the proÞt-maximizing price for an unintegrated upstream Þrm.2 The vertically integrated Þrm proÞts from the higher price because it increases the marginal cost of the Þrm’s downstream rival and results in higher downstream prices. The strategic incentive for a vertically integrated Þrm to raise the price of an intermediate good is the focus of most models in the “raising rival’s cost” literature.3 However, the success of a raising rival’s cost strategy depends not only on the vertically integrated Þrm’s incentive to increase downstream prices, but also on equilibrium conditions that determine the Þrm’s ability toraiseintermediategoodpricesandonefficiency gains from a vertically integrated market structure. We develop a model that identiÞes two opposing effects on intermediate good prices resulting from a vertical merger. A strategic effect arises from the vertically integrated Þrm’s incentive to raise the price of intermediate inputs to its downstream rivals. A confounding substitution effect is a reduction in intermediate product demand that results from the elimination of double marginalization under vertical integration. This substitution effect acts to decrease the price of the intermediate good.4

2In some situations, vertically integrated Þrms may proÞt by refusing to sell an input to downstream rivals at any price. Salinger (1991) examines incentives to foreclose rivals in a model with no product differentiation. In more general models, Hart and Tirole (1990) and Rey and Tirole (1999) show that input foreclosure can be proÞtable if an upstream monopolist cannot enter into publicly verifyable contracts with downstream Þrms. The monopolistmaynotbeabletocommititselffromßooding the market for the input and lowering its price. Vertical integration can allow the upstream Þrm to control supply and maximize proÞts. 3See, e.g., Krattenmaker and Salop (1986), Salop and Scheffman (1987), and Riordan and Salop (1995). The Þrst paper we are aware of that analyzes strategic incentives to raise rivals’ costs is Williamson (1968). 4McAfee (1999) shows that the substitution effect from a vertical merger can lower

2 In general, the effect of a vertical merger on intermediate good prices (and on the prices of downstream products) is indeterminate. We adapt the model of a vertical merger to study the effects of vertical and horizontal market structure on intermediate good prices. The model provides a theoretical foundation to explore the empirical relationships between market structure and wholesale gasoline prices that we observe in the data. We consider a market in which two Þrms merge, one of which is vertically integrated and the other is an unintegrated upstream supplier. The merger has no effect on downstream market structure and has no effect on upstream market structure if the vertically integrated Þrm was not a wholesale supplier before the merger. Nonetheless, the merger may lead to an increase in wholesale prices if the merged Þrm sells in the intermediate product market, because the merged Þrm has a strategic incentive to raise downstream rivals’ costs. Similarly, we show that mergers which increase the market share of vertically integrated Þrms are likely to increase wholesale prices if the substitution effect is small. Our model predicts that wholesale prices are positively related to the level of upstream market concentration, the market share of vertically integrated Þrms, and the degree of competition in the downstream market between integrated and unintegrated Þrms. The impact of vertical integration on consumers and on intermediate Þrms is ultimately an empirical question, yet little empirical research directly addresses the competitive effects of vertical integration. Several studies examine whether vertically integrated Þrms achieve lower costs.5 These studies do not deal with the central issue we address in this paper, namely whether vertical integration affects the prices paid by intermediate Þrms and by consumers. A few empirical papers address this question. Waterman and Weiss (1996) examine whether vertical integration of cable systems in pay cable networks (such as HBO and Showtime) has an effect on the type and number of pay networks that the cable systems carry and on the prices paid for those networks. The authors Þnd that vertically integrated cable systems are more likely to carry their affiliated networks, but they do not Þnd a statistically signiÞcant impact from vertical integration on the prices paid the equilibrium price for the intermediate good, but his model does not consider strategic incentives to raise intermediate good prices. 5 For example, Kaserman and Mayo (1991) test for economies of scope in the generation and distribution of electric power. Evans and Heckman (1984) perform a similar study for local and long distance telecommunications.

3 for network programming by the cable operator or on the prices charged to subscribers. Gertler and Cuellar (2000) explore whether vertical integration between health care providers and insurers has an effect on the prices and delivery of health care services. While these studies suggest that vertical integration can have important competitive effects, the results are confounded by the complex nature of the industries in which they occur. Pay cable networks differ in the quality of the programming that they offer. These quality differences complicate comparisons. Furthermore, vertical integration could have beneÞts that are difficult to measure, such as accelerating the launch of new networks. Similarly, quality differentials in health care delivery are signiÞcant and difficult to quantify. We empirically test the relationship between vertical structure and wholesale prices using gasoline market data. The gasoline industry is well-suited for this analysis. Gasoline of a particular grade and octane rating is a relatively homogeneous commodity, although there is a small premium for gasoline that has a well-recognized brand. IdentiÞcation of the price effects from different market structures, particularly for unbranded wholesale gasoline, is not confounded by differences in quality. There is an abundance of gasoline market data and considerable variation in the vertical and horizontal structures of gasoline markets across time and different geographies. During the mid to late 1990’s, several mergers, including the mergers of Shell and Texaco, Exxon and Mobil, and Tosco and Unocal, signiÞcantly impacted the horizontal and vertical structure of wholesale and retail gasoline markets. In addition, the presence of signiÞcant spatial differentiation at the retail (downstream) level of gasoline supply gives rise to market power and incentives for vertically integrated Þrms to raise their rivals’ costs. Gasoline markets are also interesting candidates to measure the effects of vertical integration because there is a history of allegations of anticompetitive pricing by vertically integrated Þrms that reÞne and market gasoline. Unin- tegrated gasoline retailers (often called independent marketers) have alleged that vertically integrated reÞner-marketers have engaged in price squeezes, keeping wholesale prices high and retail prices low, with the intent to mo- nopolize retail gasoline markets and raise prices. Several states have acted on these allegations and limited the ability of reÞners to own and control retail gasoline prices, while still permitting reÞners to supply wholesale gasoline to franchised retailers. Other allegations focus on regional differentials in wholesale and retail gasoline prices, sometimes contrasting prices in Los

4 Angeles and San Diego or comparing West Coast prices to prices in the Mid-Continent. Most of these claims focus on the concentration effects of upstream markets on wholesale and retail prices, but to our knowledge, no one has linked these differential pricing patterns to vertical market structures and wholesalers’ strategic incentives to raise rivals’ costs. Our empirical analysis follows two tracks. The Þrst is a study of a broad panel of U.S. metropolitan areas from January 1993 through June of 1997 to identify correlations between the average wholesale price and horizontal and vertical market structure. We Þnd that the degree of vertical integration, as well as measures of horizontal market concentration in upstream and downstream markets, are signiÞcant components of cross-sectional and intertemporal wholesale price variations. Next we turn to a speciÞcevent— the merger of Unocal and Tosco — that affected the vertical and horizontal structure of thirteen West Coast metropolitan areas. Both empirical exercises support the hypothesis that vertical market structure signiÞcantly impacts wholesale prices in ways that are consistent with our model’s predictions.

2 Raising — or lowering? — rivals’ costs

Themoststudiedexampleofraisingrivalscostisthepureverticalmerger. The analysis, such as the model in OSS, contrasts an unintegrated market structure with a (partially) vertically integrated structure. Figure 1 shows the typical unintegrated case, with two independent upstream Þrms and two independent downstream Þrms. In Figure 2, one of the upstream Þrms merges with one of the downstream Þrms. In the OSS model, the merged Þrm has an incentivetoraisethepriceoftheupstreamgood,becauseahigherupstream price causes the unintegrated downstream Þrm to raise its price, which results in higher proÞts for the integrated Þrm. The OSS model makes important assumptions that aﬀect the equilibrium responses to a vertical merger. First, the model assumes that the unintegrated upstream Þrms in Figure 1 are Nash-Bertrand competitors with constant and equal marginal costs. Therefore, the equilibrium upstream price in the unintegrated market structure is equal to the upstream Þrms’ marginal costs. As a result of this assumption, the vertical merger in Figure 2 does not generate eﬃciencies by eliminating the double marginalization that oc- curs when the upstream Þrms sell at prices that exceed their marginal costs.

5 A second important assumption in the OSS model is that the vertically integrated Þrm in Figure 2 can determine the price of the upstream good, for example, by committing to an upstream price. However, OSS do not explain why this commitment is an equilibrium outcome in the upstream market. We begin by considering a model which provides for equilibrium wholesale prices that exceed marginal costs. Hence the vertical merger will result in efficiency gains through the elimination of double marginalization. This model is consistent with many of the features of the gasoline industry. We make the following assumptions: (i) There is a homogeneous upstream good produced with zero marginal cost. In the case of gasoline, the upstream good is wholesale gasoline that is available at a terminal facility. Gasoline terminals serve a large market area. Some terminals serve a single metropolitan area, such as San Diego, Bakers- Þeld, or Sacramento. Large metropolitan areas such as the San Francisco Bay area have more than one terminal. The gasoline supplied at a terminal is a relatively homogeneous product. Independent retailers purchase unbranded gasoline at the rack, which has no brand name product differentiation, and is completely homogeneous within each octane grade. The assumption that the upstream good is produced with zero marginal cost is for convenience. The results extend easily to production with constant marginal costs. (ii) There are two upstream Þrms that act as Nash-Cournot competitors Wholesale gasoline suppliers ship gasoline from their reÞneries via pipeline to each terminal. These pipelines are often common carrier lines, and Þrms must schedule periodic shipments in anticipation of demand. Furthermore, supply at each terminal is limited by storage capacity. Unbranded gasoline is a homogeneous product and spatial differentiation at a distribution rack is minimal. Prices tend to adjust to equate demand to the available quantity. Thus the Nash-Cournot model of competition appears to be a reasonable description of competition at the wholesale level. (iii) One unit of upstream good is required to produce one unit of downstream good This assumption is appropriate to the gasoline industry. One gallon of retail gasoline requires one gallon of wholesale supply. (iv) There are two downstream Þrms that sell differentiated products with zero marginal costs. These Þrms act as Nash-Bertrand competitors.

6 Retail gasoline suppliers are spatially diﬀerentiated and post prices for their products. Again, the assumption of zero marginal cost is not signiÞ- cant.6

We consider two diﬀerent market structures, corresponding to Figures 1 and 2. In the unintegrated case, the upstream Þrms and the downstream Þrms are unrelated. In the vertical integration case, Firm 1 operates in both the upstream and the downstream markets. There is a separate upstream Firm 2 and a separate downstream Firm 2. We model competition as a two-stage game. In the Þrst stage, Firm 1 and the upstream Firm 2 choose quantities of the intermediate good that they sell to the downstream Firm 2. These quantities and the derived demands for the upstream good from the downstream Þrms determine the equilibrium price of the upstream good. In the second stage, the downstream Þrms choose retail prices taking each other’s prices as given (the Nash-Bertrand assumption) and act as a price- takers with respect to the price of the intermediate good. In the Þrst stage of the game, each Þrm takes the other’s quantity choice as given (the Nash- Cournot assumption). In the partially integrated case, Firm 1 also takes into account that its sales of the intermediate good aﬀect equilibrium retail prices. DeÞne: pi = the downstream price for sales by Þrm i =1, 2. qi = the downstream quantity sold by Þrm i =1, 2. w = the upstream price xi = sales of the upstream good by Þrm i=1,2. X= total sales of the upstream good.

2.1 The Unintegrated Case Downstream Firms: We begin with the second stage of the game, for which the price w of the upstream good is Þxed. The downstream Þrms act as Nash-Bertrand competitors, taking the upstream price w as given. Firm

6Although the retail gasoline market has features that are consistent with Nash- Bertrand competition for spatially diﬀerentiated products, we cannot rule out other con- duct such as history-dependent pricing strategies (see, e.g., Borenstein, Cameron and Gilbert, 1997). However, our results are not strongly dependent on the precise nature of competition at the retail level.

7 i chooses price pi to maximize

d π (p1,p2)=(pi w)qi(p1,p2). (1) i −

Let pi∗(w) be the equilibrium downstream prices conditional on the upstream price w. The equilibrium downstream quantities are qi(p1∗(w),p2∗(w)). One unit of the downstream good requires one unit of the upstream good. Assuming no storage, X,the total sales of the upstream good, equals q1 + q2. Thus

X = q1(p1∗(w),p2∗(w)) + q2(p1∗(w),p2∗(w)) determines the demand for the upstream good, and this can be inverted to give the inverse demand function, w(X).

Upstream Firms The upstream Þrms are Cournot-Nash competitors. They choose quantities xi (capacities) to maximize

u πi = w(xi + xj)xi. (2)

The Nash-Cournot equilibrium quantities determine a supply of the upstream good, x1∗ + x2∗ = X∗, which depends on the derived demand for the upstream good (which in turn depends on the downstream demand functions and the nature of downstream competition). The equilibrium upstream price, w∗, is the price that equates supply and demand for the upstream good:

X∗ = q1(p1∗(w∗),p2∗(w∗)) + q2(p1∗(w∗),p2∗(w∗)).

2.2 Vertical Integration Case Assume upstream Firm 1 and downstream Firm 1 merge, as indicated in Fig- ure 2. The integrated Firm 1 sells at retail and competes with the upstream Firm 2 for sales of the upstream good to the unintegrated downstream Firm 2. The integrated Þrm’s proÞtis

π1 = p1q1(p1,p2)+w(x1 + x2)x1. (3)

The unintegrated downstream Þrm’s proÞtis

d π =(p2 w)q2(p1,p2) (4) 2 −

8 and the unintegrated upstream Þrm’s proÞtis u π2 = w(x1 + x2)x2. (5) As in the unintegrated case, we consider a two-stage game. Firms are Nash-Cournot competitors in the upstream market corresponding to the Þrst stage of the game and they compete as Nash-Bertrand competitors in the second, downstream stage. Competition at the second stage results in equilibrium prices, conditional on the wholesale price, p1∗∗(w) and p2∗∗(w). In the Þrst stage, Firm 1 takes into account that its sales of the intermediate good aﬀect equilibrium retail prices in the second stage of the game. Sales of the intermediate good have to satisfy the identity

x1∗∗ + x2∗∗ = X∗∗ = q2(p1∗∗(w),p2∗∗(w)). Conditional on w, there is an inverse demand function for the upstream good, w(X), where now X = q2(p1(w),p2(w)). Note that the inverse demand for the upstream good in the vertical integration case diﬀers from the inverse demand in the unintegrated case. In the vertical integration case, X measures only sales to the unintegrated downstream Firm 2 - the demand for the wholesale product decreases.

In the Þrst stage of the game, Firm 1’s decision problem is to choose x1, taking x2 as given, to maximize

π1 = p1q1(p1,p2)+wx1 = R(p1,p2)+wx1 where p1 and p2 depend on w (but not directly on x1 and x2)andw = w(x1 + x2).

Let R1(p1,p2)=p1q1(p1(w),p2(w)). The Þrst-order condition for Firm 1 is dπ1 ∂R1 ∂p1 ∂R1 ∂p2 ∂w ∂w = + + w + x1 =0. dx1 Ã ∂p1 ∂w ∂p2 ∂w ! ∂x1 Ã ∂x1 ! The second bracketed term is the usual Þrst-order condition for the upstream Nash-Cournot competition. The Þrst term in the Þrst bracket, ∂R1 , ∂p1 is zero because the Þrms are Nash-Bertrand competitors in the second stage of the game. Thus the Þrst-order condition for Firm 1 is7

dπ1 ∂R1 ∂p2 ∂w ∂w = + w + x1 =0. (6) dx1 Ã ∂p2 ∂w ! ∂x1 Ã ∂x1 ! 7This assumes an interior solution and the usual convexity assumptions.

9 The Þrst order condition for Firm 2 is the same as in the unintegrated case, dπ2 ∂w = w + x2 =0. (7) dx2 Ã ∂x2 ! Solutions to (6) and (7) determine the supply function for the vertical integration case, x1∗∗ + x2∗∗ = X∗∗. Equating this to the demand for the upstream good, q2 = q2(p1∗∗(w),p2∗∗(w)), permits calculation of the equilibrium upstream price in the vertical integration case.

The Þrst term on the right-hand-side of (6), ∂R1 ∂p2 ∂w , is the “raising ∂p2 ∂w ∂x1 rivals’ costs” effect. By supplying less of the upstream³ good,´ Firm 1 can raise the equilibrium upstream price. This causes the unintegrated downstream Þrm to raise its price, which increases Firm 1’s proÞts. We call this the strategic effect. Figure 3 illustrates the strategic effect. This Þgure shows the Nash-Cournot reaction functions for the unintegrated and the integrated market structures. With vertical integration, Firm 1’s reaction function shifts to the left, which results in a lower equilibrium upstream quantity and a higher equilibrium upstream price. There is another effect from vertical integration which acts in the opposite direction on the upstream price. For a given upstream price, w, vertical integration lowers the unintegrated downstream Þrm’s demand for the upstream good. By eliminating double-marginalization, vertical integration results in a lower price, p1. In addition, the strategic effect identiÞed above increases the input cost to downstream Firm 2, resulting in a higher price, p2. These two effects lead to a decrease in the derived demand for the upstream good relative to the unintegrated case. McAfee (1999) identiÞed this substitution effect and showed that it can result in a lower equilibrium upstream price. Figure 4 illustrates both the strategic effect and the substitution effect. The strategic effect is a shift of the upstream supply to the left, as Firm 1 restricts output to increase the price of the intermediate good. The substitution effect is a shift of the demand function for the upstream good to the left. These two shifts can result in higher or lower equilibrium prices for the upstream good. We note, however, that the substitution effect would be mitigated to the extent that the unintegrated Þrms contract efficiently to exchange the intermediate good. Appendix A solves the system of supply and demand equations for the unintegrated and vertically integrated market structures for the special case

10 of linear downstream demand (and assuming that market exchange takes place at linear prices). The appendix shows that in the linear case, vertical integration results in a lower equilibrium price for the upstream good. The vertical merger lowers rivals’ costs!

3 Equilibrium Models of Gasoline Markets

The preceding discussion and the calculations in Appendix A are useful to frame the analysis of price effects from vertically integrated market structures in gasoline supply. However, these examples do not correspond precisely to the market data we examine. In particular, we do not observe a pure vertical merger in our data, corresponding to a movement from the market structure in Figure 1 to the market structure in Figure 2. Instead, we have data on markets that differ in the mix of vertically integrated and non-integrated (independent) reÞners and marketers. We also have data on a particular event, corresponding to the sale of upstream reÞnery assets owned by a vertically integrated reÞner-marketertoareÞner with only limited downstream assets. In this section we model the likely differentials in upstream prices corresponding to these different market structures. The model yields predictions about the price effects of vertical and horizontal market structures that can be empirically tested to assess whether the strategic incentive to raise rivals cost is a signiÞcant determinant of the intertemporal and cross-sectional variation in wholesale gasoline prices.

3.1 A mix of integrated and non-integrated ﬁrms Assume (as in Appendix A) that demands for retail gasoline are linear in prices, however we scale the size of the market of the integrated Þrm by a factor M :

q1 = M(α βp1 + γp2) − q2 = α βp2 + γp1. − The scale factor 0 elasticity to the own-elasticity of

11 demand for the two products, p ² θ = i ij −pj ²ii where ²ij = ∂ ln qi/∂ ln qj. We maintain the sequential nature of the game as discussed above. The upstream Þrms play Nash-Cournot in the Þrst stage, followed by Nash- Bertrand competition in the second stage of the game, conditional on the upstream price w. To make the model closer to reality, we assume that the independent retail segment is perfectly competitive. Therefore, in the second stage of the game, Firm 1 chooses p1 to maximize

π1 = p1q1(p1,p2)+wx1, with p2 = w, q1 = M(α βp1 + γp2), and taking w and x1 as given. The equilibrium downstream− prices are

1 α p1 = + θw (8) 2 Ãβ ! p2 = w. (9)

Sales of the upstream good must equal q2, the demand for the upstream good. Thus

1 1 2 X = q2(p1,p2)=α(1 + θ) β(1 θ )w. (10) 2 − − 2 and 1 1 w(X)= α(1 + θ) X . (11) β(1 1 θ2) 2 − − 2 · ¸ In the Þrst stage of the game, Firm 1 chooses x1 to maximize

π1 = p1q1(p1,p2)+wx1

(taking into account that w, p1, and p2 depend on x1) and the upstream Firm 2choosesx2 to maximize

u π2 = w(x1 + x2)x2.

12 The Þrst-order conditions are

∂(p1q1) ∂w ∂w + w + x1 =0 ∂w ∂x1 ∂x1 ∂w w + x2 =0. ∂x1

Thus the reaction function at the Þrst stage of the game are (for x1 > 0 and x2 > 0) 1 1 1 1 x1 = α(1 + θ) x2 Mβθp1(w), 2 2 − 2 − 2 1 1 1 x2 = α(1 + θ) x1. 2 2 − 2 1 Note the strategic effect, 2 Mβθp1(w), in Firm 1’s reaction function. If the strategic effect is sufficientlylarge,Firm1’soptimalupstreamsupplyiszero.− For x1 > 0, the total upstream supply is 2 1 1 X = α(1 + θ) Mβθp1(w). (12) 3 2 − 3 Equating upstream supply to the upstream demand given by (10), sub- stituting (8) for p1(w) and solving for the equilibrium upstream price gives 1 α 1+ 2 θ(1 + M) w = 1 2 M . (13) 3β "1 θ (1 + )# − 2 3

1 Returning to the assumption that x1 > 0,ifx1 =0,thenx2 = 2 α(1 + 1 2 θ)=X = q2(p1,p2). This implies 1 α 1+ 2 θ wmax = 1 2 . (14) 2β "1 θ # − 2 Thus the upstream price is the lesser of (13) and wmax. The derivation of the equilibrium upstream price generalizes easily to the case in which there are N 2 Þrms that supply wholesale gasoline, one of which is vertically integrated.≥ The corresponding expressions for the more general case are α 1+ 1 θ(1 + M) w = 2 (15) (N +1)β 1 1 θ2(1 + M ) − 2 N+1   13 and 1 α 1+ 2 θ wmax = 1 2 . (16) Nβ "1 θ # − 2 If N =1, the expression for the equilibrium wholesale price depends on the identity of the supplier. If the Þrm is vertically integrated, the upstream price is given by (15). If the Þrm is unintegrated, the upstream price is given by (16), and is lower than (15) if M>0. Thus the substitution of an integrated wholesale supplier for an unintegrated wholesale supplier leads to an increase in the wholesale price. We do not focus on retail prices in this analysis. However, it is easily seen that higher wholesale prices lead to higher retail prices in this model. There is a positive relationship between the factor M and the integrated Þrm’s share of total retail sales, which we derive in Appendix B. In Fig- ure 5, we show the dependence of the upstream price w on the unintegrated marketers’ share of retail sales, conditional on the substitution factor θ and the total number of (unintegrated and integrated) wholesale suppliers, N. The plateaus in Figure 5 correspond to x1 =0; the integrated Þrm does not supply the wholesale market for these values. Indeed, the data show a number of markets in which integrated Þrms do not sell gasoline at wholesale. Markets with a larger share of independent retailers have a lower equilibrium upstream price. The explanation is that the strategic effect increases with the vertically integrated market share (decreases with the share of independent retailers). An increase in the independent retailer market share has a bigger impact on the wholesale price when the substitution parameter θ is large because the strategic effect increases with θ. However, these relationships are further complicated by the fact that, in general, we should expect θ to depend on the independent retailers’ market share. These results guide the empirical examination of the effects of market structure on wholesale gasoline prices. Holding the number of upstream suppliers constant, we expect the wholesale price to increase with the market share of the vertically integrated Þrms. In addition, we expect the wholesale price to be increasing in the cross-price elasticity between integrated and unintegrated downstream Þrms. These are purely vertical market impacts. The wholesale price should be a decreasing function of N,thenumberof wholesale suppliers, whether they are integrated or unintegrated. This is a horizontal effect at the wholesale level of the market. We would also expect the wholesale price to increase with the concentration of the vertically

14 integrated suppliers. This eﬀect can be captured by the factor M,which can be construed to measure the market power of the integrated Þrm. This eﬀect has both vertical and horizontal elements. It is vertical to the extent that the higher price is driven by the strategic incentive to raise rivals’ costs and it is horizontal to the extent that upstream concentration leads to higher wholesale prices.

4 Empirical Analysis

This section empirically examines the effects of vertical integration on the wholesale price of gasoline and tests the predictions of the raising rivals’ costs model. The analysis addresses the following questions: Do prices vary with horizontal market structure in upstream markets? Does downstream market structure covary with wholesale prices? Do price patterns suggest a relationship between the degree of vertical integration and the wholesale price that is consistent with the theoretical model in Section 3? Our approach is twofold. We Þrst analyze a panel of twenty-six U.S. metropolitan areas from January 1993 through June of 1997 to identify correlations between the average wholesale price and horizontal and vertical market structure. The metropolitan are located in West Coast, Rocky Moun- tain, and Gulf Coast states. We Þnd that the degree of vertical integration, and measures of horizontal market concentration in upstream and downstream markets, are signiÞcantly correlated with wholesale prices. Next we turn to a speciÞc event — Tosco Corporation’s aquisition of Unocal’s West Coast reÞning and marketing assets — that affected the vertical and horizontal structure of thirteen West Coast metropolitan areas. We use this event to test if the differential changes in downstream market share and the degree of competition with rival independent retailers resulting from the merger were followed by changes in wholesale prices as predicted by our model. Both empirical exercises support the hypothesis that vertical market structure sig- niÞcantly impacts wholesale prices in a manner that is consistent with the theoretical model developed in the previous section.

4.0.1 Data Description: Market Structure and Wholesale Prices Retail Census data, available annually from Whitney Leigh Corporation for each of the twenty-six metropolitan areas used in this analysis, provide de-

15 tailed information on the characteristics of every retail gasoline outlet in each metropolitan area. The retail census also includes the ownership and delivery type for each station. This variable determines each staiton’s vertical relationship (if any) with the upstream reÞner. Hence, the Retail Censuses allow us to examine the degree of vertical integration of each of the upstream reÞners selling unbranded gasoline to rival unintegrated retailers in each metropolitan area. We identify stations that are owned by a reÞner as vertically integrated stations, whether or not the retail price is set by the reÞner or by a residual claimant. This assumption is consistent with the results in Hastings (2000), which show no signiÞcant diﬀerence in the pricing behavior between a reÞner’s directly operated stations and those operated by dealers.8 Lundberg Wholesale Price Reports provide semi-monthly, average, unbranded wholesale prices for each metropolitan area during the time period considered. Unbranded gasoline is gasoline that is supplied at wholesale at a distribution “rack” (a supply terminal) with no restrictions on its point of sale. These average price data are coupled with data from Oil Price Infor- mation Service on the names of the companies supplying at each distribution rack during each time period. Thus the compiled data sets provide detailed information on the average unbranded wholesale price, the companies who are supplying unbranded gasoline at each rack, the downstream market share of each of these companies in each metropolitan area (including if they have no downstream market share at all), as well as the market share of unintegrated retail marketers who purchase unbranded gasoline at the rack and compete with the integrated downstream Þrms.

Summary Statistics: Table 1 provides summary statistics of the market structure variables of interest. The table shows that there is a great deal of variationwithinthepaneldataset.Muchofthisvariationiscross-sectional. However, there is substantial inter-temporal variation within metropolitan areas, largely due to entry, exit and various mergers that occurred in the mid-1990’s. For example, the independent retail chain, Stop-N-Go, was purchased by Ultramar Diamond Shamrock, a reÞner that was integrated in 8This result is also consistent with eﬃcient contracting. The reÞner can set a station- speciÞc wholesale price based on the station’s demand elasticity, and a station-speciÞc lease rate with a volume discount to extract the retail rents. Hence the reÞner’s incentives to raise rival retailers’ costs is similar for reÞner-owned and reÞner-supplied stations.

16 Gulf Coast markets, but not in others. This merger affected markets in Petroleum Administration Defense Districts (PADDs) 3 and 4. Tosco, an unintegrated reÞner, purchased Brittish Petroleum’s PaciÞcNorthWestre- Þning and marketing assets. Tosco also purchased Circle K, an unintegrated retailer in a number of Southwest gasoline markets. ARCO, an integrated reÞner, purchased Thrifty, an unintegrated retailer in Southern California. These mergers, and other events, caused considerable variation in the horizontal and vertical market structure variables across the city-time units of observation. Table 2 presents the sample correlation between the market structure variables of interest. The correlation coefficients in Table 2 indicate that horizontal market structure is not highly correlated with vertical market structure. In other words, market structures in the sample are not simply two extreme types: markets with very few unintegrated wholesalers and very high degree of vertical integration, or markets with a very unconcentrated wholesale market and a small degree of vertical integration. For example, in the third quarter of 1995, Albuquerque, New Mexico had a very concentrated upstream market and a very unconcentrated downstream market. The distribution rack had only 3 reÞners supplying wholesale product — one vertically integrated with a downstream market share of about Þfteen percent, and two unintegrated reÞners. However, the independent retail market share was a relatively high twenty-seven percent. In the same quarter, New Orleans, Louisiana had relatively unconcentrated upstream and downstream markets, with ten reÞners supplying at the rack, eight of which were unintegrated, and an independent retail market share of about twenty-three percent. Seat- tle, Washington had concentrated upstream and downstream markets, with one integrated reÞner and two unintegrated reÞners supplying at the distribution rack, and an independent retail market share of only eight percent. Across all of the markets and time periods summarized in Table 2, the correlation between the market share of independent retailers and the number of unintegrated suppliers is positive, but small. In addition, the correlation between the degree of vertical integration, downstream market share, and the upstreammarketconcentrationofbothintegratedandunintegratedreÞners is negative, but also small. Table 2 illustrates that there is sufficient variation to independently identify the wholesale price effects of both horizontal upstream market structure and vertical market structure. Table 3 presents the data in cell means, which allow us to partition average

17 prices according to market structure characteristics. In Table 3, the data are grouped by combinations of high and low numbers of vertically integrated suppliers, high and low numbers of unintegrated suppliers, high and low market shares of independent retail marketers, and high and low average downstream market shares of the vertically integrated suppliers. For each variable, high and low are determined by above and below the median value in Table 2. The dependent variable is the average price of unbranded wholesale gasoline in each metropolitan area less the average spot price for crude oil.9 Although prices are available semi-monthly, they are averaged each quarter for this analysis, because the right-hand-side variables change in discrete jumps, but are fairly constant relative to the wholesale price of gasoline over shorter time periods. Table 3 shows the average price (wholesale margin over crude oil price) forobservationsineachcell.Forexamplecell1,inthetopleftcornercorre- sponds to city-quarter observations with few integrated wholesale suppliers, a small average downstream market share for the integrated suppliers, few unintegrated wholesale suppliers, and a small market share for independent retailers. Cell 5 (row 1 and column 2) shows the average price for city- quarters with the highest values for all market structures. This cell is for city-quarters with few integrated wholesale suppliers, a large average downstream market share for the integrated suppliers, few unintegrated wholesale suppliers, and a small market share for independent retailers. It is also the cell with the highest wholesale price. Cell 12 (row 4 and column 3) is the cell for the most unconcentrated city-quarter markets. This cell has many upstream suppliers of both types, a large market share for independents, and a low average downstream market share for the integrated reÞners. This cell has the lowest average price. Moreover, the average price in cell 5 is roughly twicetheaveragepriceincell12. In addition, conditioned on the other factors, the average price of gasoline in cells with fewer integrated suppliers is higher than cells with many integrated suppliers. Likewise, conditioned on other factors, cells with fewer unintegrated suppliers have higher average prices than cells with many unintegrated suppliers. These two stylized facts in the data are consistent with the traditional price eﬀects of horizontal concentration. As the number of upstream competitors increases, the wholesale price decreases.

9The crude price is the spot price at Cushing provided by the Energy Information Administration.

18 Qrwlfh wkdw wkh wzr grzqvwuhdp pdunhw vwuxfwxuh yduldeohv dovr kdyh d v|vwhpdwlf uhodwlrqvkls zlwk wkh fhoo phdqv lq Wdeoh 61 Frqglwlrqhg rq wkh rwkhu idfwruv/ pdunhwv zlwk odujh vkduhv ri lqghshqghqw uhwdlohuv kdyh orzhu zkrohvdoh sulfhv wkdq pdunhwv zlwk vpdoo vkduhv ri lqghshqghqw uhwdlohuv1 Ixuwkhupruh/ wkh zkrohvdoh sulfh lv kljkhu lq pdunhwv zkhuh wkh zkrohvdoh vxssolhuv kdyh d juhdwhu dyhudjh grzqvwuhdp pdunhw vkduh/ frqglwlrqhg rq wkh rwkhu idfwruv1 Wklv glhuhqfh lv prvw surqrxqfhg lq pdunhwv vwuxfwxuhv iru zklfk wkh lqwhjudwhg zkrohvdoh upv kdyh wkh juhdwhvw lqfhqwlyh dqg delo0 lw| wr udlvh grzqvwuhdp ulydov frvwv1 Frpsdulqj fhoov 4 dqg 8 +urz 4/ froxpq 4 dqg 5 uhvshfwlyho|,/ zh vhh wkdw lq pdunhwv zkhuh wkhuh duh d ihz grzq0 vwuhdp ulydov/ dqg zkhuh wkhuh duh ihz xsvwuhdp vxssolhuv hlwkhu lqwhjudwhg ru xqlqwhjudwhg/ wkh glhuhqfh ehwzhhq wkh dyhudjh sulfh lq fhoov zkhuh xs0 vwuhdp upv kdyh d odujh dyhudjh grzqvwuhdp pdunhw vkduh dqg wkrvh zkhuh wkh| kdyh d vpdoo rqh lv odujh dqg vljql fdqw1 Wkh glhuhqfh lq wkh fhoo dyhudjhv lv 51:7 fhqwv shu jdoorq/ dqg lv vwdwlvwlfdoo| vljql fdqw dw wkh vhyhq shufhqw ohyho zlwk dq I ydoxh ri 61571 Wkhvh fhoo phdq sdwwhuqv duh jhqhudoo| frqvlvwhqw zlwk wkh wkhruhwlfdo prgho1 Qrwlfh wkdw prylqj iurp fhoo 4 wr fhoo 6 +froxpq 4/ urz 4 dqg urz 6 uhvshf0 wlyho|,/ zklfk fruuhvsrqgv wr dgglqj pruh xqlqwhjudwhg vxssolhuv lq d pdunhw zkhuh wkhuh duh ihz lqghshqghqwv dqg ihz lqwhjudwhg vxssolhuv zlwk d odujh dyhudjh grzqvwuhdp pdunhw vkduh/ ghfuhdvhv wkh dyhudjh sulfh e| 5154 fhqwv/ zklfk lv vljql fdqw dw wkh whq shufhqw ohyho1 Wkhvh fhoo frpsdulvrqv vxjjhvw wkdw d pryh wr d vpdoo dyhudjh grzqvwuhdp pdunhw vkduh iru wkh lqwhjudwhg vxssolhu kroglqj rwkhu idfwruv frqvwdqw/ +wkh pryhphqw iurp fhoo 8 wr fhoo 4,/ kdv d vlplodu hhfw rq wkh fhoo phdq dv lqfuhdvlqj wkh qxpehu ri xqlqwhjudwhg vxssolhuv +wkh pryhphqw iurp fhoo 4 wr fhoo 6,1 Wkhvh ydoxhv vxjjhvw wkdw wkh ghjuhh ri yhuwlfdo lqwhjudwlrq lv dq lpsruwdqw ghwhuplqdqw ri zkrohvdoh sulfhv shukdsv dv lpsruwdqw dv krul}rqwdo xsvwuhdp frqfhqwudwlrq1 D pryhphqw iurp d fhoo lq Wdeoh 6 zlwk ihz lqghshqghqw pdunhwhuv wr rqh zlwk pdq| lqghshqghqw pdunhwhuv |lhogv d orzhu fhoo phdq/ h{fhsw zkhq prylqj iurp fhoo 8 wr fhoo 9/ ru iurp fhoo 48 wr fhoo 491 Irfxvlqj rq fhoov 8 dqg 9/ zkhuh wkhuh duh ihz xsvwuhdp vxssolhuv dqg wkh lqwhjudwhg vxssolhuv kdyh d odujh pdunhw vkduh/ wkh wkhru| vxjjhvwv wkdw wkh lqfhqwlyh ri lqwhjudwhg upv wr udlvh wkh zkrohvdoh sulfh ghfuhdvhv zlwk wkh lqghshqghqw uhwdlo pdunhw vkduh dqg lqfuhdvhv zlwk w1 Krzhyhu/ wkhuh lv qr fkdqjh lq wkh fhoo phdq ehwzhhq fhoov 8 dqg 91 Wklv pd| eh gxh wr wkh idfw wkdw wkh lqghshqghqw uhwdlo pdunhw vkduh dqg wkh furvv0hodvwlflw| sdudphwhu w duh doprvw fhuwdlqo|

4< uhodwhg/ zlwk w dq lqfuhdvlqj ixqfwlrq ri wkh lqghshqghqw uhwdlo pdunhw vkduh1 Vlqfh uhwdlo jdvrolqh vwdwlrqv duh jhrjudsklfdoo| glhuhqwldwhg surgxfwv/ dv wkh qxpehu ri lqghshqghqw uhwdlohuv lqfuhdvhv/ wkhlu dyhudjh glvwdqfh wr dq lqwhjudwhg uhwdlo vwdwlrq pd| ghfuhdvh/ ohdglqj wr dq lqfuhdvh lq wkh furvv0 hodvwlflw| sdudphwhu w iru jhrjudsklfdoo| glhuhqwldwhg surgxfwv1 Wkxv wkhvh wzr hhfwv pd| eh frqirxqglqj lq vrph fhoo phdqv/ exw zh zloo eh deoh wr vhsdudwh wkhp lq wkh ghwdlohg dqdo|vlv ri wkh Wrvfr0Xqrfdo phujhu lq wkh qh{w vhfwlrq1 Wdeoh 7 suhvhqwv d uhjuhvvlrq fruuhodwlrq ehwzhhq wkh frqwlqxrxv ydoxhv ri wkh pdunhw vwuxfwxuh yduldeohv dqg wkh zkrohvdoh sulfhv1 Wkh uhjuhvvlrq fruuhodwlrq uh hfwv wkh uhodwlrqvklsv hylghqw lq wkh fhoo phdq suhvhqwdwlrq/ zkhuh wkh uhjuhvvruv* ydoxhv zhuh glylghg lqwr glvfuhwh fdwhjrulhv1 Xsvwuhdp frqfhqwudwlrq lv srvlwlyho| fruuhodwhg zlwk sulfh/ wkh pdunhw vkduh ri lqgh0 shqghqwv lv qhjdwlyho| fruuhodwhg zlwk sulfh/ dqg wkh dyhudjh pdunhw vkduh ri wkh yhuwlfdoo| lqwhjudwhg vxssolhuv frydulhv srvlwlyho| zlwk zkrohvdoh sulfh1

714 Ghwdlohg Sdqho Uhjuhvvlrq= Wkh suhfhglqj vhfwlrq rhuv lpsruwdqw lqvljkwv lqwr zkrohvdoh sulfh yduld0 wlrq dqg lwv uhodwlrqvkls wr yhuwlfdo dqg krul}rqwdo pdunhw vwuxfwxuh1 Wkh fhoo phdqv vxjjhvw wkdw wkh vwudwhjlf lqfhqwlyh iru yhuwlfdoo| lqwhjudwhg upv wr udlvh ulydov zkrohvdoh frvwv pd| eh d vljql fdqw ghwhuplqdqw ri furvv0 vhfwlrqdo dqg lqwhuwhpsrudo sulfh yduldwlrq1 Wkh phujhu ri Wrvfr dqg Xqrfdo doorzv d pruh suhflvh lghqwl fdwlrq ri wkh uhodwlrqvkls ehwzhhq wkh ghjuhh ri grzqvwuhdp lqwhjudwlrq dqg zkrohvdoh sulfhv 0 wkh yhuwlfdo frpsrqhqw lq krul}rqwdo phujhuv1 Wkh phujhu fdxvhg glvfuhwh fkdqjhv lq Wrvfr*v grzq0 vwuhdp pdunhw vkduh lq wkluwhhq Zhvw Frdvw phwursrolwdq duhdv1 Zh h{sorlw wklv yduldwlrq wr whvw li wkhvh fkdqjhv lq grzqvwuhdp pdunhw vwuxfwxuh zhuh iroorzhg e| fkdqjhv lq Wrvfr*v zkrohvdoh jdvrolqh sulfh dv suhglfwhg e| rxu prgho/ frqwuroolqj iru frvwv dqg rwkhu lpsruwdqw pdunhw vwuxfwxuh yduldeohv1 Pruhryhu/ lq wklv dqdo|vlv zh duh deoh wr phdvxuh wkh jhrjudsklf sur{lp0 lw| ri wkh lqwhjudwhg vwdwlrqv wr ulydo lqghshqghqwv/ jlylqj d phdvxuh ri wkh furvv0sulfh hodvwlflw| ehwzhhq wkh lqwhjudwhg dqg xqlqwhjudwhg grzqvwuhdp upv1

53 71414 Ghwdlov ri wkh Phujhu= Lq Qryhpehu ri 4<<9 Wrvfr dqg Xqrfdo dqqrxqfhg wkh sursrvhg vdoh ri Xq0 rfdo*v Zhvw Frdvw uh qlqj dqg pdunhwlqj dvvhwv wr Wrvfr1 Wklv lqfoxghg doo ri Xqrfdo*v uh qhulhv rq wkh Zhvw Frdvw/ dqg doo ri wkhlu uhwdlo rxwohwv/ lqfoxglqj wkh Xqlrq :9 orjr/ lq doo Zhvw Frdvw phwursrolwdq duhdv lqfoxg0 lqj wkrvh lq Dul}rqd dqg Qhydgd1 Wkh sxufkdvh zdv frpsohwhg lq Dsulo ri 4<<:/ zkhq Wrvfr r!fldoo| wrrn frqwuro ri wkh Xqrfdo dvvhwv dqg wkhlu rs0 hudwlrq1 Xqrfdo*v grzqvwuhdp uhwdlohuv zhuh doprvw h{foxvlyho| lqwhjudwhg1 Wkhlu grzqvwuhdp pdunhw vkduh ydulhg iurp wzr shufhqw wr hljkwhhq shufhqw ri wkh wrwdo fhqvxv ri uhwdlo vwdwlrqv lq wkh phwursrolwdq duhdv frqvlghuhg1 Lq vrph phwursrolwdq duhdv/ Xqrfdo vrog vrph dprxqw ri xqeudqghg jdvrolqh dw wkh glvwulexwlrq udfn/ exw lq rwkhuv wkh| glg qrw frpshwh lq wkh xqeudqghg zkrohvdoh pdunhw1 Wdeoh 8 vkrzv wkh pdunhwv wkdw zhuh dhfwhg e| wkh sxufkdvh/ zlwk wkh fruuhvsrqglqj fkdqjhv lq grzqvwuhdp pdunhw vkduh dqg xsvwuhdp frqfhqwudwlrq1 Ehiruh wkh dftxlvlwlrq ri Xqrfdo*v Zhvw Frdvw grzqvwuhdp dvvhwv/ Wrvfr kdg vrph grzqvwuhdp pdunhw suhvhqfh lq d ihz ri wkh flwlhv frqvlghuhg1 Wkhlu uhwdlo pdunhw vkduh ydulhg iurp }hur wr dssur{lpdwho| iruw| shufhqw1 Lq wkh ihz flwlhv zkhuh Wrvfr kdg d vljql fdqw grzqvwuhdp pdunhw vkduh/ Xqrfdo w|slfdoo| glg qrw kdyh d vljql fdqw qxpehu ri grzqvwuhdp rxwohwv1 Lq wklv uhvshfw/ wkh dvvhwv ri Wrvfr dqg Xqrfdo zhuh frpsohphqwdu| lq hdfk phwursrolwdq duhd +wkh dftxlvlwlrq zdv sulpdulo| yhuwlfdo,/ khqfh wkh phujhu pd| qrw kdyh udlvhg wudglwlrqdo dqwlwuxvw frqfhuqv ryhu lqfuhdvhv lq krul}rq0 wdo frqfhqwudwlrq lq xsvwuhdp ru grzqvwuhdp pdunhwv1 Wrvfr*v grzqvwuhdp dvvhwv lqfoxghg wkh Eulwlvk Shwurohxp vwdwlrqv dqg wkh ES eudqg +Wrvfr sxufkdvhg ES*v Qruwkzhvw uh qlqj dqg pdunhwlqj dvvhwv lq 4<<6, dqg wkh Flufoh N frqyhqlhqfh vwruh dqg jdvrolqh vwdwlrq fkdlq/ zklfk wkh| dftxluhg dw wkh hqg ri Pd|/ 4<<9143 ES vwdwlrqv zhuh prvwo| lq wkh Sdfl f Qruwkzhvw phwursrolwdq duhdv +Vhdwwoh dqg Sruwodqg lq rxu gdwd,/ zlwk d vpdoo pdunhw 43Wkh ES dftxlvlwlrq dhfwhg ihzhu pdunhwv wkdq glg wkh Xqrfdo0Wrvfr phujhu1 Lq dgglwlrq/ Wrvfr glg qrw vxsso| zkrohvdoh surgxfw lq pdq| ri wkh dhfwhg pdunhwv sulru wr wkh phujhu zlwk ES1 Khqfh/ pdq| pdunhwv gr qrw surylgh d ehiruh dqg diwhu frpsdulvrq rq Wrvfr*v zkrohvdoh sulfh1 Wkh yhuwlfdo phujhu zlwk Flufoh N surylghv dq lqwhuhvwlqj rsruwxqlw| wr orrn dw udvlqj ulydo*v frvw lq d sxuho| yhuwlfdo phujhu/ krzhyhu wkh wlplqj frlqflghg zlwk wkh uhtxluhphqw ri Fdoliruqld Dlu Uhvrufhv Erdug +FDUE, uhirupxodwhg jdvrolqh1 Khqfh/ hyhq wkrxjk zh irxqg d glvfuhwh mxps lq zkrohvdoh sulfh lq dhfwhg pdunhwv diwhu wkh yhuwlfdo phujhu/ wklv fdqqrw eh lqghshqghqwo| lghqwl hg iurp d flw|0 vshfl f FDUE uhjxodwlrq hhfw1

54 vkduh lq Vdq Iudqflvfr/ dqg Flufoh N vwdwlrqv zhuh orfdwhg prvwo| lq Dul}rqd/ zlwk d ihz vwdwlrqv lq Qhydgd dqg Vrxwkhuq Fdoliruqld1 Wrvfr kdg d ihz uh0 wdlo rxwohwv uhodwlyh wr uh qlqj fdsdflw| lq prvw duhdv/ dqg vrog zkrohvdoh jdvrolqh wr lqghshqghqw pdunhwhuv lq doo ri wkh wkluwhhq phwursrolwdq duhdv wkdw duh frqvlghuhg lq wklv hpslulfdo dqdo|vlv1 Wkh dftxlvlwlrq ri Xqrfdo*v grzqvwuhdp dvvhwv e| Wrvfr fkdqjhg wkh grzqvwuhdp pdunhw vkduh iru Wrvfr e| ydu|lqj ghjuhhv dfurvv wkh Zhvw Frdvw phwursrolwdq duhdv1 Vrph pdunhwv zhuh sudfwlfdoo| xqdhfwhg e| wkh phujhu/ zkloh lq rwkhuv Wrvfr h{shulhqfhg dq lqfuhdvh lq grzqvwuhdp pdunhw vkduh ri xs wr vl{whhq shufhqw1 Zh hpslulfdoo| whvw li wkh glvfuhwh fkdqjhv lq wkh ghjuhhv ri yhuwlfdo lqwhjudwlrq zhuh iroorzhg e| fkdqjhv lq Wrvfr*v sulfhv iru xqeudqghg zkrohvdoh jdvrolqh/ frpphqvxudwh zlwk wkh suhglfwlrqv ri rxu prgho1 Lq sduwlfxodu/ zh xvh wkh yduldwlrq lq wkh ghjuhh ri yhuwlfdo lqwhjud0 wlrq uhvxowlqj iurp wkh phujhu wr whvw li Wrvfr*v zkrohvdoh sulfh lv lqfuhdvlqj lq lwv grzqvwuhdp pdunhw vkduh dqg lqfuhdvlqj lq lwv ghjuhh ri frpshwlwlrq zlwk lqghshqghqw uhwdlohuv/ frqwuroolqj iru yduldwlrq lq frvwv dqg krul}rqwdo pdunhw vwuxfwxuh yduldeohv1

71415 Gdwd= Iru wklv hpslulfdo dqdo|vlv zh xvh wkh Uhwdlo Fhqvxv gdwd ghvfulehg lq wkh suhylrxv vhfwlrq/ frxsohg zlwk ghwdlohg zkrohvdoh sulfh gdwd iurp Rlo Sulfh Lqirupdwlrq Vhuylfh1 Wkh uhwdlo fhqvxv gdwd uhsruwhg lq Wdeoh 9 surylgh d phdvxuh ri wkh fkdqjh lq Wrvfr*v grzqvwuhdp pdunhw vkduh wkdw uhvxowhg lq hdfk phwursrolwdq duhd dv uhvxow ri wkh dftxlvlwlrq ri Xqrfdo*v uhwdlo rxwohwv1 Wkh ghwdlohg zkrohvdoh udfn sulfh gdwd iurp Rlo Sulfh Lqirupdwlrq Vhuylfh surylgh zhhno| dyhudjh zkrohvdoh jdvrolqh sulfhv e| vxssolhu dqg phwursrol0 wdq duhd1 Wkh sulfh gdwd duh iru wkh shulrg ri Mxo| 4<<9 wkurxjk Ghfhpehu ri 4<<;/ surylglqj zhhno| revhuydwlrqv rq Wrvfr*v sulfh iru xqeudqghg jdvrolqh dw hdfk glvwulexwlrq udfn iru derxw rqh |hdu ehiruh/ dqg d |hdu dqg d kdoi diwhu wkh hyhqw ri lqwhuhvw1 Xvlqj wkhvh gdwd/ zh frqvwuxfw wzr pdlq yduldeohv wr lghqwli| wkh hhfwv ri udlvlqj ulydov frvw lqfhqwlyhv rq Wrvfr*v xqeudqghg zkrohvdoh sulfh ri jdvr0 olqh1 Wkh uvw lv wkh ghshqghqw yduldeoh= wkh sulfh ri zkrohvdoh jdvrolqh1 Ehfdxvh rxu irfxv lv rq vwudwhjlf sulfh hhfwv/ zh glhuhqfh zkrohvdoh sulfhv wr dffrxqw iru idfwruv vxfk dv lqsxw frvwv dqg lqyhqwrulhv1 Vshfl fdoo|/ zh vxewudfw wkh sulfh ri zkrohvdoh jdvrolqh lq Skrhql{ iurp wkh zkrohvdoh sulfh

55 wlph vhulhv lq hdfk phwursrolwdq duhd1 Zh fkrrvh Wrvfr*v zkrohvdoh sulfh ri jdvrolqh lq Skrhql{ dv dq dssur{lpdwlrq ri frvw yduldeohv ehfdxvh Skrhql{ h{shulhqfhg qr fkdqjh lq xsvwuhdp ru grzqvwuhdp pdunhw vwuxfwxuh gxulqj wkh wlph shulrg frqvlghuhg dqg/ lq sduwlfxodu/ zdv qrw lpsdfwhg e| wkh Xqrfdo dftxlvlwlrq144 Wkh Skrhql{ zkrohvdoh sulfhv surylgh d ehwwhu dssur{lpdwlrq ri wkh sdqrso| ri frvw idfwruv wkdw lpsdfw zkrohvdoh sulfhv lq lqwhufrqqhfwhg Zhvw Frdvw pdunhwv wkdq gr fuxgh rlo sulfhv145 Lq wklv vhfwlrq zh dovr frqvwuxfw d yduldeoh/ zklfk zh fdoo grzqvwuhdp pdunhw frqwdfw/ wkdw surylghv d ehwwhu dssur{lpdwlrq wkdq uhwdlo vkduh dorqh iru wkh frpshwlwlrq wkdw h{lvwv lq grzqvwuhdp pdunhwv1 Wkh wkhru| ghvfulehg lq Vhfwlrq 6 vkrzv wkdw wkh vwudwhjlf lqfhqwlyh wr udlvh ulydov* frvwv ghshqgv rq wkh ghpdqg sdudphwhu w dv zhoo dv rq wkh pdunhw vkduh ri lqghshqghqw +wkdw lv/ ulydo, uhwdlohuv1 Wkh wkhru| suhglfwv wkdw wkh htxloleulxp zkrohvdoh sulfh lqfuhdvhv zlwk w dqg ghfuhdvhv zlwk wkh pdunhw vkduh ri lqghshqghqw uhwdlohuv1 Wkxv/ lq dgglwlrq wr grzqvwuhdp pdunhw vkduh dv dq h{sodqdwru| yduldeoh/ wkh hpslulfdo dqdo|vlv vkrxog dwwhpsw wr hvwlpdwh wkh pdjqlwxgh ri wkh sdudphwhu w1 Uhwdlo jdvrolqh rxwohwv duh jhrjudsklfdoo| glhuhqwldwhg146 Lq pdunhwv zkhuh wkh pdunhw vkduh ri lqghshqghqw uhwdlohuv lv yhu| vpdoo/ wkh lqwhudfwlrq ehwzhhq lqghshqghqw dqg yhuwlfdoo| lqwhjudwhg uhwdlohuv/ dv sdudphwhul}hg e| wkh idfwru w lq wkh wkhruhwlfdo dqdo|vlv/ lv dovr olnho| wr eh vpdoo ehfdxvh/ rq dyhudjh/ lqghshqghqw uhwdlohuv duh glvwdqw iurp yhuwlfdoo| lqwhjudwhg uhwdlohuv1 Dowhuqdwlyho|/ li prvw Xqrfdo vwdwlrqv kdg dq lqghshq0 ghqw uhwdlohu dv d forvh frpshwlwru ehiruh wkh dftxlvlwlrq/ d vpdoo lqfuhdvh lq grzqvwuhdp pdunhw vkduh iurp wkh Xqrfdo sxufkdvh frxog kdyh d vljql fdqw lpsdfw rq Wrvfr*v zkrohvdoh sulfh vwudwhj|1 Wklv kljk ghjuhh ri lqwhudfwlrq lv pruh olnho| zkhq wkh pdunhw vkduh ri lqghshqghqwv lv odujh1 Xvlqj Uhwdlo Fhq0 vxv gdwd/ zh frqvwuxfw d yduldeoh wkdw phdvxuhv wkh lqfuhdvh lq grzqvwuhdp pdunhw frpshwlwlrq zlwk lqghshqghqwv uhvxowlqj iurp wkh Xqrfdo sxufkdvh1 Wklv lv grqh e| zhljkwlqj wkh lqfuhdvh lq wkh grzqvwuhdp pdunhw vkduh gxh 44Rqo| 618 shufhqw ri wkh vwdwlrqv lq Skrhql{ duh Lqghshqghqw pdunhwhuv1 Vr hyhq wkrxjk Wrvfr*v pdunhw vkduh lqfuhdvhg e| 5 shufhqw/ lwv grzqvwuhdp frqwdfw zlwk lqghshqghqw uhwdlohuv glg qrw lqfuhdvh dw doo dv d uhvxow ri wkh Xqrfdo dtfxlvlwlrq1 45Zh xvhg fuxgh rlo sulfhv wr dgmxvw iru frvw idfwruv lq wkh eurdg sdqho dqdo|vlv glvfxvvhg lq wkh suhylrxv vhfwlrq ehfdxvh qr flw| pdunhw kdg dq xqfkdqjhg pdunhw vwuxfwxuh ryhu wkh hqwluh shulrg fryhuhg lq wkh eurdg sdqho1 Wkh uhjuhvvlrq uhvxowv suhvhqwhg lq Wdeoh 9 gr qrw fkdqjh vljql fdqwo| li dgg wkh sulfh lq Skrhql{ dv d uhjuhvvru lqvwhdg ri vxewudfwlqj lw iurp wkh ohiw kdqg vlgh1 Wkh frh!fhqw rq wkh Sulfh lq Skrhql{ lv 31<861 46Vhh Vodgh +4<;9/ 4<<5,/ Eruhqvwhlq/ Fdphurq/ dqg Jloehuw +4<<:,/ dqg Kdvwlqjv +5333,1

56 wr wkh dftxlvlwlrq ri Xqrfdo uhwdlo rxwohwv e| wkh shufhqw ri wkrvh rxwohwv wkdw duh jhrjudsklfdoo| orfdwhg zlwklq rqh ploh ri dq lqghshqghqw uhwdlohu1 Fuh0 dwlqj wklv yduldeoh iru Wrvfr*v uhwdlo rxwohwv erwk ehiruh dqg diwhu wkh phujhu doorzv xv wr whvw li Wrvfr*v zkrohvdoh sulfh lv srvlwlyho| lpsdfwhg e| wklv phd0 vxuh ri grzqvwuhdp pdunhw frpshwlwlrq zlwk lqghshqghqwv/ dv suhglfwhg e| rxu prgho1

71416 Hpslulfdo Vshfl fdwlrq dqg Uhvxowv=

Zh hvwlpdwh wkh iroorzlqj uhjuhvvlrq wr hvwlpdwh yduldwlrq lq Wrvfr*v glhu0 hqfhg zkrohvdoh sulfh dv d ixqfwlrq ri grzqvwuhdp pdunhw vkduh/ grzqvwuhdp frqwdfw zlwk lqghshqghqw uhwdlohuv/ xsvwuhdp frqfhqwudwlrq/ dqg wkh pdunhw vkduh ri lqghshqghqw uhwdlohuv1 Wkh yduldwlrq lq Wrvfr*v grzqvwuhdp dvvhwv uhvxowlqj iurp wkh phujhu doorz xv wr vhsdudwho| lghqwli| wkh sulfh hhfwv iurp wkhvh idfwruv dqg rwkhu flw|0ohyho {hg hhfwv wkdw frydu| zlwk sulfhv1 Wkh uhjuhvvlrq lv vshfl hg dv iroorzv=

R| ' > n k n q_| n | n Bo| n 0|

0| ' 40c|3 n 1| Zkhuh=

R| ' Wrvfr*v zhhno| dyhudjh sulfh ri xqeudqghg zkrohvdoh jdvrolqh lq flw| ohvv wkh sulfh lq Skrhql{ lq zhhn | k ' flw|0vshfl f {hg hhfw _| ' grzqvwuhdp pdunhw frqwdfw zlwk ulydo lqghshqghqwv lq flw| lq zhhn | | ' wkh qxpehu ri uh qhuv vhoolqj xqeudqghg jdvrolqh lq flw| lq zhhn | o| ' wkh shufhqw ri ulydo lqghshqghqw uhwdlohuv lq flw| lq zhhn | 0| ' dxwruhjuhvvlyh huuru frpsrqhqw 1| ' zklwh qrlvh huuru whup Wkh qxpehu ri uh qhuv vhoolqj xqeudqghg jdvrolqh lv d yduldeoh wkdw frxqwv vxssolhuv zkr srvwhg sulfhv dw hdfk glvwulexwlrq udfn lq hdfk wlph shulrg1 Wkhuh lv vrph lqwhuwhpsrudo yduldwlrq lq wklv uhjuhvvru1 Vrph zdv gxh wr wkh srvw0phujhu h{lw ri Xqrfdo iurp wkh pdunhwv zkhuh lw vxssolhg xqeudqghg jdvrolqh1 Rwkhu yduldwlrq frphv iurp shulrglf hqwu| ru h{lw e| uh qhuv dw hdfk udfn1 Lq 4<<;/ diwhu Vkhoo dqg Wh{dfr phujhg/ Wh{dfr fhdvhg vhoolqj xq0 eudqghg jdvrolqh lq pdq| pdunhwv zkhuh wkh frpelqhg upv kdg d vljql fdqw grzqvwuhdp pdunhw vkduh1 Wklv surylghv ixuwkhu yduldwlrq lq wkh qxpehu ri xsvwuhdp frpshwlwruv ryhu rxu vdpsoh shulrg1

57 Wkh huuru frpsrqhqw lv jlyhq dq dxwruhjuhvvlyh vwuxfwxuh wr fdswxuh wkh g|qdplf hhfw ri jdvrolqh sulfhv1 D Glfnh|0Ixoohu whvw iru d xqlw urrw uhmhfwhg wkh k|srwkhvlvri d xqlw urrw lq wkh zkrohvdoh sulfh wlph vhulhv iru hdfk flw| lqfoxghg lq wkh uhjuhvvlrq1 Lq dgglwlrq/ wkh vshfl fdwlrq ri dq dxwruhjuhvvlyh huuru vwuxfwxuh lv qrw uhmhfwhg lq idyru fdswxulqj wkh g|qdplf hhfw zlwk d odjjhg sulfh ydoxh rq wkh uljkw0kdqg vlgh1 Wkh huuru vwuxfwxuh lv dovr jlyhq d flw|0vshfl f frpsrqhqw1 D Kdxvpdq whvw grhv qrw uhmhfw wkh k|srwkhvlv wkdw wkh flw|0vshfl f huuru frpsrqhqw lv xqfruuhodwhg zlwk wkh uhjuhvvruv/ khqfh zh fdqqrw uhmhfw wkh udqgrp hhfwv prgho lq idyru ri d {hg0hhfwv vshfl fdwlrq147 Wdeoh 9 suhvhqw uhjuhvvlrq uhvxowv1 Wkh uvw vshfl fdwlrq lqfoxghv flw|0 vshfl f {hg0hhfwv dqg wkh grzqvwuhdp pdunhw frqwdfw yduldeoh/ zklfk phd0 vxuhv w lq wkh wkhruhwlfdo prgho1 Wkh yduldwlrq lq grzqvwuhdp pdunhw frqwdfw uhvxowlqj iurp wkh phujhu doorzv xv wr vhsdudwho| lghqwli| wkh sulfh hhfw ri wklv yduldeoh iurp wkh flw|0ohyho {hg hhfwv dqg wkh rwkhu fryduldwhv lq0 foxghg lq wkh uhjuhvvlrq1 Wkh uhjuhvvlrq uhvxowv lqglfdwh wkdw d kljkhu ghjuhh ri grzqvwuhdp frqwdfw zlwk lqghshqghqw uhwdlohuv lv dvvrfldwhg zlwk d kljkhu sulfh iru xqeudqghg jdvrolqh1 Wkh vsrw hvwlpdwh lpsolhv wkdw iru hyhu| rqh shufhqw lqfuhdvh lq wkh shufhqw ri vwdwlrqv lq gluhfw frpshwlwlrq zlwk d ulydo lqghshqghqw uhwdlohu/ wkh zhhno| dyhudjh xqeudqghg jdvrolqh sulfh ulvhv e| 3177 fhqwv shu jdoorq1 Frqvlvwhqw zlwk rxu prgho/ dq lqfuhdvh lq grzqvwuhdp pdunhw frqwdfw zlwk ulydo lqghshqghqw upv lqfuhdvhv wkh vwudwhjlf lqfhqwlyh wr udlvh ulydo*v frvwv1 Wkh hvwlpdwhv lpso| wkdw lq Orv Dqjhohv/ iru h{dpsoh/ zkhuh wkh dftxlvlwlrq ri Xqrfdo*v uhwdlo dvvhwv lqfuhdvhg wkh qxpehu ri lq0 whjudwhg vwdwlrqv wkdw frpshwh zlwk ulydo lqghshqghqwv dqg khqfh lqfuhdvhg Wrvfr*v lqfhqwlyh wr udlvh ulydo*v frvwv/ wkh hvwlpdwhg sulfh lqfuhdvh lv 61: fhqwv shu jdoorq1 Lq wkh {hg0hhfwv vshfl fdwlrq lq Wdeoh :/ wkh frh!flhqwv rq wkh qxp0 ehu ri xsvwuhdp frpshwlwruv dqg wkh pdunhw vkduh ri lqghshqghqwv duh qrw vljql fdqwo| glhuhqw iurp }hur1 Krzhyhu/ wkh frh!flhqw rq wkh qxpehu ri

47 3 4 Kdxvpdq*v p ydoxh lv p @ t Ydu+t, t/ zkhuh t @ eIH eUH dqg Ydu+t,@ Ydu+eIH, Ydu+eUH,1 Wkh qxoo k|srwkhvlv lv wkdw H+dlm[l,@3yhuvxv wkh dowhuqd0 wlyh wkdw lw lv qrw htxdo wr }hur1 Xqghu wkh qxoo k|srwkhvlv/ wkh vwdwlvwlf lv glvwulexwhg fkl0vtxduhg zlwk N ghjuhhv ri iuhhgrp1 Li wkh qxoo lv uhmhfwhg/ wkh udqgrp0hhfwv vshf0 l fdwlrq lv lqfruuhfw1 Udqgrp0hhfwv dvvxphv wkdw wkh glvwulexwlrq ri wkh flw|0vshfl f huuru frpsrqhqw/ frqglwlrqhg rq wkh uhjuhvvruv/ lv wkh vdph dfurvv doo flwlhv1 Il{hg0hhfwv hvwlpdwhv wkh phdq ri wklv frpsrqhqw dqg grhv qrw uhtxluh lw wr eh }hur1 Li H+dlm[l, 9@3 wkh Udqgrp0hhfwv hvwlpdwru lv lqfrqvlvwhqw1

58 xsvwuhdp frpshwlwruv lv vljql fdqw lq wkh udqgrp0hhfwv vshfl fdwlrq suh0 vhqwhg lq wkh vhfrqg froxpq/ zklfk dvvxphv wkdw wkh frqglwlrqdo phdq ri wkh flw|0vshfl f huuru frpsrqhqw lv wkh vdph dfurvv doo flwlhv1 Wkh frh!flhqw lv qhjdwlyh/ dqg zrxog lqglfdwh wkdw rqh ohvv xsvwuhdp frpshwlwru zrxog ohdg wr d 31736 fhqw lqfuhdvh lq dyhudjh zkrohvdoh sulfh1 Wklv sdudphwhu hvwlpdwh djuhhv zlwk wkhru|/ dqg uhsuhvhqwv wkh wudglwlrqdo krul}rqwdo frpsrqhqw wr d phujhu1 Phwursrolwdq duhdv zkhuh wkhuh zdv dq lqfuhdvh lq pdunhw wkh grzq0 vwuhdp frqwdfw zlwk ulydo lqghshqghqwv/ dqg zkhuh wkhuh zdv dq lqfuhdvh lq frqfhqwudwlrq uhvxowlqj iurp wkh phujhu/ zrxog h{shulhqfh kljkhu zkrohvdoh sulfhv iurp erwk wkh yhuwlfdo dqg krul}rqwdo hhfwv ri wkh phujhu1 Zkhq wkh flw|0ohyho {hg hhfwv duh lqfoxghg/ wkh frh!flhqw rq xsvwuhdp frpshwlwruv ehfrphv lqvljql fdqw/ suredeo| ehfdxvh wkhuh lv qrw vx!flhqw yduldwlrq lq wklv uhjuhvvru wr suhflvho| lghqwli| lw vhsdudwho| iurp wkh {hg0hhfw1 Wkh frh!flhqw rq wkh pdunhw vkduh ri lqghshqghqw uhwdlo pdunhwhuv lv qrw vwdwlv0 wlfdoo| vljql fdqw lq hlwkhu huuru frpsrqhqw vshfl fdwlrq1

Wkh uhvxowv lq Wdeoh 9 lpso| wkdw dq lqwhjudwhg uh qhu*v sulfh iru xq0 eudqghg zkrohvdoh jdvrolqh lv dq lqfuhdvlqj ixqfwlrq ri lwv frpshwlwlrq zlwk ulydo lqghshqghqw uhwdlohuv1 Wkh frh!flhqw rq wkh ghjuhh ri grzqvwuhdp frp0 shwlwlrq zlwk ulydov fdswxuhv erwk wkh hhfwv ri wkh Wrvfr dftxlvlwlrq dqg dq| suh0phujhu ydudlwlrq dfurvv phwursrolwdq duhdv1 Zh fdq irfxv rq wkh dftxl0 vlwlrq*v hhfw e| h{dplqlqj krz fkdqjhv lq flw|0dyhudjh sulfhv ehiruh dqg diwhu wkh dftxlvlwlrq ydu| zlwk wkh lqfuhdvh lq grzqvwuhdp frqwdfw zlwk lq0 ghshqghqw uhwdlohuv1 Li/ lqghhg/ wkh sulfh fkdqjhv duh uhodwhg wr grzqvwuhdp pdunhw frqwdfw/ wkh hvwlpdwhg frh!flhqw vkrxog eh wkh vdph dv wkh frh!flhqw suhvhqwhg lq Wdeoh 91

Iljxuh 9 suhvhqwv d vfdwwhu sorw ri wkh hvwlpdwhg sulfh lqfuhdvh lq hdfk phwursrolwdq duhd iroorzlqj wkh Xqrfdo sxufkdvh/ djdlqvw wkh lqfuhdvh lq grzqvwuhdp frpshwlwlrq zlwk ulydo lqghshqghqw uhwdlohuv1 Wkhvh hvwlpdwhg sulfh lqfuhdvhv duh wkhq sorwwhg djdlqvw wkh lqfuhdvh lq grzqvwuhdp pdunhw frqwdfw zlwk lqghshqghqw uhwdlohuv1 Wdeoh : suhvhqwv wkh vorsh frh!flhqw iru wkh ROV uhjuhvvlrq ri wkh hvwlpdwhg sulfh fkdqjh rq wkh fkdqjh lq grzq0 vwuhdp frqwdfw zlwk lqghshqghqwv1 Wkh frh!flhqw rq Grzqvwuhdp Pdunhw Frqwdfw lv 316967 dqg lv vljql fdqw dw wkh rqh0shufhqw ohyho1 Wklv frh!flhqw lv wkh vdph dv wkh rqh lq Froxpq 5 ri Wdeoh 91 Khqfh/ wkh uhvxowv djdlq vxj0 jhvw wkdw dq lqwhjudwhg uh qhu*v xqeudqghg zkrohvdoh sulfh lv dq lqfuhdvlqj ixqfwlrq ri wkh ghjuhh ri frpshwlwlrq zlwk lqghshqghqw uhwdlohuv/ dv suhglfwhg

59 e| rxu Udlvlqj Ulydov* Frvw prgho1

Wdeoh ; suhvhqwv uhvxowv zkhuh wkh grzqvwuhdp pdunhw frqwdfw yduldeoh lv uhsodfhg zlwk wkh yhuwlfdoo| lqwhjudwhg grzqvwuhdp pdunhw vkduh1 Wkh uhjuhvvlrq uhvxowv lqglfdwh d srvlwlyh uhodwlrqvkls ehwzhhq grzqvwuhdp pdunhw vkduh dqg wkh xqeudqghg zkrohvdoh sulfh1 Wkh frh!flhqw lpsolhv wkdw iru hyhu| 4 shufhqw lqfuhdvh lq grzqvwuhdp pdunhw vkduh/ Wrvfr*v sulfh ulvhv e| 314<; fhqwv shu jdoorq1 Iru Vdq Mrvh/ wklv lpsolhv d 51<7 fhqw d jdoorq lqfuhdvh lq wkh sulfh ri xqeudqghg jdvrolqh uhvxowlqj iurp wkh dftxlvlwlrq ri Xqrfdo*v uhwdlo rxwohwv1

8 Frqfoxvlrq

Wkh lpsdfw ri yhuwlfdo pdunhw vwuxfwxuh rq zkrohvdoh +dqg uhwdlo, sulfhv lv lq jhqhudo gl!fxow wr suhglfw1 Wkhuh duh wkhruhwlfdo uhdvrqv zk| upv zlwk pdunhw srzhu wkdw rshudwh dw erwk wkh zkrohvdoh dqg uhwdlo ohyhov ri d pdu0 nhw zrxog kdyh dq lqfhqwlyh wr udlvh wkh zkrohvdoh sulfh +ru uhixvh wr vxsso| dw zkrohvdoh, lq rughu wr glvdgydqwdjh wkhlu grzqvwuhdp frpshwlwruv1 Krz0 hyhu/ wkhvh vwudwhjlf lqfhqwlyhv fdq eh rvhw e| fkdqjhv lq wkh ghpdqg iru wkh xsvwuhdp jrrg wkdw dffrpsdq| fkdqjhv lq yhuwlfdo pdunhw vwuxfwxuh1 Wkh edodqfh ri wkhvh wzr hhfwv ghshqgv rq idfwxdo flufxpvwdqfhv lq hdfk pdunhw1 Lqghhg/ zh qg wkdw d sxuho| yhuwlfdo phujhu +d iuhtxhqwo| flwhg h{dpsoh ri udlvlqj ulydov* frvwv, fdq ohdg wr d uhgxfwlrq lq zkrohvdoh sulfhv1 Pruhryhu/ fkdqjhv lq yhuwlfdo pdunhw vwuxfwxuh fdq kdyh frpsolfdwhg lpsdfwv rq rwkhu yduldeohv/ vxfk dv wkh furvv0hodvwlflw| ehwzhhq lqghshqghqw dqg yhuwlfdoo| lq0 whjudwhg upv/ zklfk dhfw wkh lqfhqwlyhv iru yhuwlfdoo| lqwhjudwhg upv wr udlvh wkhlu ulydov* frvwv1 Wkxv lqyhvwljdwlrq ri wkh lpsdfwv ri yhuwlfdo pdunhw vwuxfwxuhv uhtxluhv d fduhixo hpslulfdo dqdo|vlv wkdw lv jxlghg e| wkhruhwlfdo suhglfwlrqv1 Zh dqdo|}h wkh h{shfwhg zkrohvdoh sulfh hhfwv ri glhuhqw yhuwlfdo pdu0 nhw vwuxfwxuhv iru wkh fdvh ri jdvrolqh pdunhwv1 Jdvrolqh pdunhwv kdyh wkh dwwudfwlyh ihdwxuh wkdw wkh surgxfw lv krprjhqhrxv/ dqg dffxudwh dqg dexq0 gdqw gdwd duh dydlodeoh wr hvwlpdwh vwuxfwxudo lpsdfwv1 Zh qg hylghqfh lq d eurdg sdqho wkdw yhuwlfdo lqwhjudwlrq pdwwhuv iru xsvwuhdp uhwdlo sulfhv dqg wkdw zkrohvdoh sulfhv whqg wr eh kljkhu lq pdunhwv zlwk odujh yhuwlfdoo| lqwhjudwhg upv1 Wklv qglqj lv frqvlvwhqw zlwk wkh vwudwhjlf lqfhqwlyh dqg delolw| ri yhuwlfdoo| lqwhjudwhg upv wr udlvh lqsxw frvwv wr grzqvwuhdp ulydov1

5: Zh ixuwkhu whvw wklv fruuhodwlrq zlwk d vshfl f hyhqw wkdw fdxvhg glvfuhwh dqg glhuhqwldo fkdqjhv lq wkh ghjuhh ri yhuwlfdo lqwhjudwlrq lq jdvrolqh pdunhwv dfurvv Zhvw Frdvw phwursrolwdq duhdv1 Wkh qglqjv iurp wklv pruh ghwdlohg dqdo|vlv duh frqvlvwhqw zlwk wkh fruuhodwlrqv suhvhqwhg lq wkh eurdg sdqho ri X1V1 phwursrolwdq duhdv1 Zh qg wkdw wkh ghjuhh ri yhuwlfdo lqwhjudwlrq kdv vljql fdqw dqg srvlwlyh lpsdfwv rq lqwhjudwhg upv* zkrohvdoh sulfhv1 Pruhryhu/ wkh lqfhqwlyh wr udlvh sulfh lv dovr srvlwlyho| fruuhodwhg zlwk wkh jhrjudsklf sur{lplw| ri lqwhjudwhg vwdwlrqv wr ulydo lqghshqghqwv/ lqglfdwlqj wkdw wkh juhdwhu wkh ghjuhh ri frpshwlwlrq/ ru furvv0sulfh hodvwlflw|/ ehwzhhq lqwhjudwhg uhwdlohuv dqg ulydo lqghshqghqw uhwdlohuv/ wkh juhdwhu wkh lqwhjudwhg up*v lqfhqwlyh wr udlvh ulydov* zkrohvdoh frvwv1

5; Uhihuhqfhv

^4` Eduurq Mrkq P1 dqg Mrkq U1 Xpehfn/ Wkh Hhfwv ri Glhuhqw Frq0 wudfwxdo Duudqjhphqwv= Wkh Fdvh ri Uhwdlo Jdvrolqh Pdunhwv Mrxuqdo ri Odz dqg Hfrqrplfv/ Yro1 5: qr1 5 +Rfwrehu 4<;7, ss1 646065<1

^5` Eruhqvwhlq/ Vhyhulq/ Vhoolqj Frvwv dqg Vzlwfklqj Frvwv= H{sodlqlqj Uhwdlo Jdvrolqh Pdujlqv UDQG Mrxuqdo ri Hfrqrplfv/Yro155qr16 +Dxwxpq 4<<4, ss1 687069<1

^6` Eruhqvwhlq/ Vhyhulq dqg Dqguhd Vkhsdug/ Vwlfn| Sulfhv/ Lqyhqwrulhv/ dqg Pdunhw Srzhu lq Zkrohvdoh Jdvrolqh Pdunhwv1 Qdwlrqdo Exuhdx ri Hfrqrplf Uhvhdufk Zrunlqj Sdshu &879;/ 4<<9e1

^7` Eruhqvwhlq/ Vhyhulq dqg Dqguhd Vkhsdug/ G|qdplf Sulflqj lq Uhwdlo Jdvrolqh Pdunhwv1 UDQG Mrxuqdo ri Hfrqrplfv/ Yro1 5: qr1 6 +Dxwxpq 4<<9, ss175<05841

^8` Eruhqvwhlq/ Vhyhulq Frolq Fdphurq dqg Ulfkdug Jloehuw/ Gr Jdvrolqh Sulfhv Uhvsrqg Dv|pphwulfdoo| wr Fuxgh Rlo sulfh FkdqjhvB Txduwhuo| Mrxuqdo ri Hfrqrplfv/ Yro1 445 qr14 +Iheuxdu| 4<<:, ss1 639066<1

^9` Eruhqvwhlq/ Vhyhulq dqg Ulfkdug Jloehuw/ Xqfoh Vdp dw wkh Jdv Sxps= Fdxvhv dqg Frqvhtxhqfhv ri Uhjxodwlqj Jdvrolqh Glvwulexwlrq1 Uhjx0 odwlrq/ Yro1 49 qr1 5 +Vsulqj 4<<6, ss1 960:81

^:` Fxhoodu/ Dolvrq dqg Sdxo Jhuwohu/ Sk|vlfdq0Krvslwdo Yhuwlfdo Lqwhjud0 wlrq/ Xqlyhuvlw| ri Fdoliruqld dw Ehunhoh| zrunlqj sdshu +5333,1

^;` Hfrqrplghv/ Qlfkrodv/ Wkh Lqfhqwlyh iru Qrq0Sulfh Glvfulplqdwlrq e| dq Lqsxw Prqrsrolvw/ Lqwhuqdwlrqdo Mrxuqdo ri Lqgxvwuldo Rujdql}dwlrq/ Yro1 49 qr16/ +Vsulqj 4<<6, ss1 5:405;71

^<` Hydqv/ Gdylg V1 dqg Mdphv M1 Khfnpdq/ D Whvw iru Vxedgglwlylw| ri wkh Frvw Ixqfwlrq zlwk dq Dssolfdwlrq wr wkh Ehoo V|vwhp / Dphulfdq Hfrqrplf Uhylhz/ Yro1 :7 qr1 7 +Vhswhpehu 4<;7, ss1 94809561

5< ^43` Kduw/ Rolyhu dqg Mhdq Wluroh/ Yhuwlfdo Phujhuv dqg Pdunhw Iruhfor0 vxuh/ Eurrnlqjv Sdshuv rq Hfrqrplf Dfwlylw| 0 Plfurhfrqrplfv/ +4<<3, ss1 53705:91

^44` Kdvwlqjv/ Mxvwlqh/Yhuwlfdo Uhodwlrqvklsv dqg Frpshwlwlrq lq Uhwdlo Jdvrolqh Pdunhwv= Hpslulfdo Hylghqfh iurp Frqwudfw Fkdqjhv lq Vrxwk0 huq Fdoliruqld/ Xqlyhuvlw| ri Fdoliruqld Hqhuj| Lqvwlwxwh Srzhu Zrun0 lqj Sdshu/ +5334, SZS03:81

^45` Mrvnrz/ Sdxo O1/ Yhuwlfdo Lqwhjudwlrq dqg Orqj0Whup Frqwudfwv= Wkh Fdvh ri Frdo0Exuqlqj Hohfwulf Jhqhudwlqj Sodqwv/ Mrxuqdo ri Odz/ Hfr0 qrplfv/ dqg Rujdql}dwlrq/ Yro1 4 +4<;8, ss1 660;31

^46` Ndvhupdq/ Gdylg O1 dqg Mrkq Z1 Pd|r/ Wkh Phdvxuhphqw ri Yhuwlfdo Hfrqrplhv dqg wkh H!flhqw Vwuxfwxuh ri wkh Hohfwulf Xwlolw| Exvlqhvv/ Mrxuqdo ri Lqgxvwuldo Hfrqrplfv/ Yro1 6< qr1 8/ ss1 7;608351

^47` Nudwwhqpdnhu dqg Vwhyhq F1 Vdors/ Dqwlfrpshwlwlyh H{foxvlrq= Udlv0 lqj Ulydov* Frvwv wr Dfklhyh Srzhu Ryhu Sulfh/ \doh Odz Mrxuqdo/ Yro1 <9 qr1 5 +Ghfhpehu 4<;9, ss1 53<05<61

^48` Nduuhqeurfn/ Mhuh| G1/ Wkh Ehkdylru ri Uhwdlo Jdvrolqh Sulfhv= V|p0 phwulf ru QrwB Ihghudo Uhvhuyh Edqn ri Vw1 Ohzlv Uhylhz/ Yro1 :6 qr1 7 +Mxo|2Dxjxvw 4<<4, ss14<05<1

^49` PfDihh/ U1 Suhvwrq/ Wkh Hhfwv ri Yhuwlfdo Lqwhjudwlrq rq Frpshwlqj Lqsxw Vxssolhuv/ Hfrqrplf Uhylhz +Ihghudo Uhvhuyh Edqn ri Fohyh0 odqg,/ Yro1 68 qr1 4 +Iluvw Txduwhu 4<<<, ss1 50;1

^4:` Rugryhu/ Mdqx}v D1/ Jduwk Vdorqhu/ dqg Vwhyhq F/ Vdors/ Htxloleulxp Yhuwlfdo Iruhforvxuh/ Dphulfdq Hfrqrplf Uhylhz/ Yro1 ;3 qr1 4 +Pdufk 4<<3, ss1 45:04751

^4;` Uh|/ Sdwulfn dqg Mhdq Wluroh/ D Sulphu rq Iruhforvxuh/ zrunlqj sdshu +4<<:,1

^4<` Uh|/ Sdwulfn dqg Mrvhsk H1 Vwljolw}/ Wkh Uroh ri H{foxvlyh Whuulwrulhv lq Surgxfhuv* Frpshwlwlrq/ Udqg Mrxuqdo ri Hfrqrplfv/Yro159qr16 +Dxwxpq 4<<8, ss1 76407841

63 ^53` Ulrugdq/ Plfkdho K1 dqg Vwhyhq F1 Vdors/ Hydoxdwlqj Yhuwlfdo Phujhuv= D Srvw0Fklfdjr Dssurdfk/ Dqwlwuxvw Odz Mrxuqdo/ Yro1 96 Lvvxh 5/ ss1 846089;1

^54` Vdolqjhu/ Plfkdho D1/ Yhuwlfdo Phujhuv dqg Pdunhw Iruhforvxuh/ Txdu0 whuo| Mrxuqdo ri Hfrqrplfv/ Yro1 436 qr15 +Pd| 4<;;,/ ss1 67806891

^55` Vdolqjhu/ Plfkdho D1/ Yhuwlfdo Phujhuv lq Pxow0Surgxfw Lqgxvwulhv dqg Hgjhzruwk*v Sdudgr{ ri Wd{dwlrq/ Mrxuqdo ri Lqgxvwuldo Hfrqrplfv/ Yro1 6< qr1 8 +Vhswhpehu 4<<4, ss1 87808891

^56` Vdors/ Vwhyhq F1 dqg Gdylg W1 Vfkhpdq/ Frvw0Udlvlqj Vwudwhjlhv/ Mrxuqdo ri Lqgxvwuldo Hfrqrplfv/ Yro1 69 qr1 4/ +Vhswhpehu 4<;:, ss1 4<0671

^57` Vkhsdug/ Dqguhd/ Frqwudfwxdo Irup/ Uhwdlo Sulfh/ dqg Dvvhw Fkdudf0 whulvwlfv1 UDQG Mrxuqdo ri Hfrqrplfv/ Yro1 57 qr1 4 +Vsulqj 4<<6, ss1 8;0::1

^58` Vkhsdug/ Dqguhd/ Sulflqj Ehkdylru dqg Frqwudfwv lq Uhwdlo Pdunhwv Dphulfdq Hfrqrplf Uhylhz/ Yro1 ;3 qr1 5 +Pd| 4<<3, ss175:07641

^59` Vkhsdug/ Dqguhd/ Sulfh Glvfulplqdwlrq dqg Uhwdlo Frq jxudwlrq Mrxuqdo ri Srolwlfdo Hfrqrp|/ Yro1 << qr1 4 +Iheuxdu| 4<<4, ss1 630861

^5:` Vkhsdug/ Dqguhd/ Frqwudfwxdo Irup/ Uhwdlo Sulfh/ dqg Dvvhw Fkdudf0 whulvwlfv lq Jdvrolqh Uhwdlolqj1 Udqg Mrxuqdo ri Hfrqrplfv/Yro157qr1 4 +Vsulqj4<<6, ss18;0::1

^5;` Vodgh/ Pdujduhw H1/ H{rjhqhlw| Whvwv ri Pdunhw Erxqgdulhv Dssolhg wr Shwurohxp Surgxfwv1 Mrxuqdo ri Lqgxvwuldo Hfrqrplfv/Yro167qr16 +Pdufk 4<;9, ss1 5<406361

^5<` Vodgh/ Pdujduhw H1/ Frqmhfwxuhv/ Ilup Fkdudfwhulvwlfv/ dqg Pdunhw Vwuxfwxuh= Dq Hpslulfdo Dvvhvvphqw1 Lqwhuqdwlrqdo Mrxuqdo ri Lqgxv0 wuldo Rujdql}dwlrq/ Yro1 7 qr1 7 +Ghfhpehu 4<;9, ss1 67:069<1

^63` Vodgh/ Pdujduhw H1/ Lqwhu up Ulydou| lq d Uhshdwhg Jdph= Dq Hpslu0 lfdo Whvw ri Wdflw Frooxvlrq1 Mrxuqdo ri Lqgxvwuldo Hfrqrplfv/ Yro1 68 qr1 7 +Mxqh 4<;:, ss17<<08481

64 ^64` Vodgh/ Pdujduhw H1/ Ydqfrxyhu*v Jdvrolqh0Sulfh Zduv= Dq Hpslul0 fdo H{huflvh lq Xqfryhulqj Vxshujdph Vwudwhjlhv1 Uhylhz ri Hfrqrplf Vwxglhv/ Yro1 8< qr1 5 +Dsulo 4<<5, ss1 58:0:91

^65` Vodgh/ Pdujduhw H1/ Hpslulfdo Jdphv= Wkh Roljrsro| Fdvh1 Fdqdgldq Mrxuqdo ri Hfrqrplfv/ Yro1 5; qr1 5 +Pd| 4<<8, ss1 69;07351

^66` Vodgh/ Pdujduhw H1/ Pxowlwdvn Djhqf| dqg Frqwudfw Fkrlfh= Dq Hp0 slulfdo H{sorudwlrq1 Lqwhuqdwlrqdo Hfrqrplf Uhylhz/Yro16:qr15+Pd| 4<<9, ss1 79807;91

^67` Vodgh/ Pdujduhw H1/ Vwudwhjlf Prwlyhv iur Yhuwlfdo Vhsdudwlrq= Hyl0 ghqfh iurp Uhwdlo Jdvrolqh Pdunhwv Mrxuqdo ri Odz/ Hfrqrplfv/ dqg Rujdql}dwlrq/ Yro1 47 qr1 4 +Dsulo 4<<;, ss1 ;704461

^68` Wluroh/ Mhdq/ Wkh Wkhru| ri Lqgxvwuldo Rujdql}dwlrq/ Wkh PLW Suhvv= Fdpeulgjh/ Pdvvdfkxvhwwv/ 4<;;1

^69` Zdwhupdq/ Gdylg dqg Dqguhz D1 Zhlvv/ Wkh Hhfwv ri Yhuwlfdo Lqwh0 judwlrq Ehwzhhq Fdeoh Whohylvlrq V|vwhpv dqg Sd| Fdeoh Qhwzrunv/ Mrxuqdo ri Hfrqrphwulfv/ Yro1 :5 qr1 4/5 +Pd|2Mxqh 4<<9, ss1 68:06<81

^6:` Zlooldpvrq/ Rolyhu H1/ Pdunhwv dqg Klhudufklhv= Dqdo|vlv dqg Dqwlwuxvw Lpsolfdwlrqv/ Iuhh Suhvv/ 4<:81

^6;` Zlooldpvrq/ Rolyhu/ Zdjh Udwhv dv d Eduulhu wr Hqwu|= wkh Shqqlqjwrq Fdvh lq Shuvshfwlyh/ Txduwhuo| Mrxuqdo ri Hfrqrplfv/ +Iheuxdu| 4<9;, ss1 ;804491

65 Dsshqgl{ D

Dv lq wkh pdlq wh{w/ zh prgho d wzr vwdjh jdph zlwk Qdvk0Frxuqrw frpshwlwlrq lq wkh uvw +xsvwuhdp, vwdjh dqg Qdvk0Ehuwudqg frpshwlwlrq lq wkh vhfrqg +grzqvwuhdp, vwdjh1 Zh frqvlghu wzr glhuhqw pdunhw vwuxfwxuhv/ fruuhvsrqglqj wr Iljxuhv 4 dqg 51 Lq wkh xqlqwhjudwhg fdvh/ wkh xsvwuhdp upv dqg wkh grzqvwuhdp upv duh xquhodwhg1 Lq wkh yhuwlfdo lqwhjudwlrq fdvh/ Ilup 4 rshudwhv lq erwk wkh xsvwuhdp dqg wkh grzqvwuhdp pdunhwv1 Wkhuh lv d vhsdudwh xsvwuhdp Ilup 5 dqg d vhsdudwh grzqvwuhdp Ilup 51 Ehjlqqlqj zlwk wkh grzqvwuhdp uhwdlo pdunhw/ zh dvvxph wkdw qdo gh0 pdqgv duh olqhdu lq sulfhv/

^ ' k qR n R2

^2 ' k qR2 n R zlwk q:f dqg :f Iru vwdelolw| +gh qh,/ zh dvvxph q: Gh qh w ' G w phdvxuhv wkh udwlr ri wkh furvv0hodvwlflw| wr wkh rzq0hodvwlflw| ri q ghpdqg iru wkh grzqvwuhdp surgxfwv1

814 Wkh Xqlqwhjudwhg Fdvh Grzqvwuhdp Ilupv Zh ehjlq zlwk wkh vhfrqg vwdjh ri wkh jdph/ iru zklfk wkh sulfh ri wkh xsvwuhdp jrrg lv {hg1 Wkh grzqvwuhdp upv dfw dv Qdvk0Ehuwudqg frpshwlwruv/ wdnlqj wkh xsvwuhdp sulfh dv jlyhq1 Ilup fkrrvhv sulfh R wr pd{lpl}h _ Z ERcR2'ER ^ERcR2 Wkh uhdfwlrq ixqfwlrqv duh k R ' n n wR +4:, 2q 2 2 2 k R ' n n wR +4;, 2 2q 2 2 Wkh htxloleulxp sulfhv duh

k n q R ' R ' +4<, 2 2qE w 2 66 dqg wkh htxloleulxp ghpdqgv iru wkh surgxfwv ri wkh grzqvwuhdp upv duh

k qE w ^ ' ^ ' +53, 2 2E w 2 Htxdwlrq +53, gh qhv dq lqyhuvh ghpdqg ixqfwlrq iru wkh xsvwuhdp jrrg1 Rqh xqlw ri wkh grzqvwuhdp jrrg uhtxluhv rqh xqlw ri wkh xsvwuhdp jrrg1 Ohw f eh wkh wrwdo vdohv ri wkh xsvwuhdp jrrg +dqg dvvxph qr vwrudjh,1 Wkhq f ' ^ n ^2/ dqg xvlqj +53, zh kdyh wkh lqyhuvh ghpdqg iru wkh xsvwuhdp jrrg= k E wf Ef' 2 +54, qE w Xsvwuhdp Ilupv Wkh xsvwuhdp upv duh Frxuqrw0Qdvk frpshwlwruv1 Wkh| fkrrvh txdq0 wlwlhv % +fdsdflwlhv, wr pd{lpl}h

Z ' E% n %% Wkh uhdfwlrq ixqfwlrqv duh k % ' % 2E w 2 2 2 k % ' % 2 2E w 2 2 dqg wkh htxloleulxp txdqwlwlhv +fdsdflwlhv, duh k % ' % ' dqg 2 E w 2 2k f ' % n % ' +55, 2 E w 2

Iurp +55,/ +4<, dqg wkh lghqwlw| f ' ^EREcR2E n^2EREcR2E/ zh fdq frpsxwh wkh htxloleulxp xsvwuhdp sulfh +zkhuh XL vwdqgv iru wkh xqlqwhjudwhg fdvh,1 Wkh ghpdqg iru wkh xsvwuhdp jrrg lv ^ n ^ ' dk qE wo +56, 2 E w 2

67 dqg wkh frqglwlrq wkdw vxsso| htxdov ghpdqg lpsolhv k LU ' +57, qE w Wkh htxloleulxp grzqvwuhdp sulfhv duh % &% & 2k w RLU ' RLU ' e +58, 2 q E wE w 2

Qrwh wkdw li ' f/ fruuhvsrqglqj wr vhsdudwh pdunhwv/ wkhq ^ ' ^2 ' k cwkh xsvwuhdp sulfh lv LU ' k / dqg wkh htxloleulxp sulfhv duh RLU ' q RLU ' 2k Qrwh wkdw wkh grzqvwuhdp sulfhv h{fhhg wkh sur w0pd{lpl}lqj 2 q sulfh/ R6 ' k Wklv lv wkh uhvxow ri grxeoh0pdujlqdol}dwlrq/ zklfk lv prghu0 2q dwhg vrphzkdw e| Frxuqrw frpshwlwlrq xsvwuhdp1 Htxloleulxp grzqvwuhdp sulfhv duh dq lqfuhdvlqj ixqfwlrq ri w iru w Wkh grzqvwuhdp surgxfwv duh vwudwhjlf frpsohphqwv zlwk Qdvk0Ehuwudqg frpshwlwlrq/ khqfh sulfhv ulvh dv wkh furvv0hodvwlflw| sdudphwhu lqfuhdvhv1

815 Yhuwlfdo Lqwhjudwlrq Fdvh Dvvxph xsvwuhdp Ilup 4 dqg grzqvwuhdp Ilup 4 phujh/ dv lqglfdwhg lq Ilj0 xuh 41 Wkh lqwhjudwhg Ilup 4 vhoov dw uhwdlo dqg frpshwhv zlwk wkh xsvwuhdp Ilup 5 iru vdohv ri wkh xsvwuhdp jrrg wr wkh xqlqwhjudwhg grzqvwuhdp Ilup 51 Dv ehiruh/ zh dvvxph Qdvk0Frxuqrw frpshwlwlrq xsvwuhdp dqg Qdvk0 Ehuwudqg frpshwlwlrq grzqvwuhdp1 Gh qh=

% ' vdohv ri wkh xsvwuhdp jrrg e| Ilup wr grzqvwuhdp Ilup 51 TU ' sulfh ri wkh xsvwuhdp jrrg +zkhuh YL vwdqgv iru wkh yhuwlfdo lqwhjudwlrq fdvh,1 Wkh lqwhjudwhg up*v sur w lv

Z ' R^ERcR2nE% n %2% Wkh xqlqwhjudwhg grzqvwuhdp up*v sur w lv

_ Z2 ' ER2 ^2ERcR2 dqg wkh xqlqwhjudwhg xsvwuhdp up*v sur w lv

Z2 ' E% n %2%2

68 Dv lq wkh xqlqwhjudwhg fdvh/ zh frqvlghu d wzr0vwdjh jdph1 Ilupv duh Qdvk0Frxuqrw frpshwlwruv lq wkh xsvwuhdp pdunhw fruuhvsrqglqj wr wkh uvw vwdjh ri wkh jdph dqg wkh| frpshwh dv Qdvk0Ehuwudqg frpshwlwruv lq wkh vhfrqg/ grzqvwuhdp vwdjh1 Lq wkh uvw vwdjh/ Ilup 4 wdnhv lqwr dffrxqw wkdw lwv vdohv ri wkh lqwhuphgldwh jrrg dhfw htxloleulxp uhwdlo sulfhv lq wkh vhfrqg vwdjh ri wkh jdph1 Wkh uhdfwlrq ixqfwlrqv lq wkh vhfrqg vwdjh ri wkh jdph duh k R ' n wR c 2q 2 2 k R ' n n wR 2 2q 2 2 Frqglwlrqdo rq /htxloleulxp uhwdlo sulfhv duh % & % & k w RTU ' E n wn +59, E w2 2q 2 e % e & % & k RTU ' E n wn 2 E w2 2q 2 2 e Vdohv ri wkh lqwhuphgldwh jrrg kdyh wr vdwlvi| wkh lghqwlw|

% n %2 ' f ' ^2ERcR2'k qR2 n R Qrwh wkdw wkh lqyhuvh ghpdqg iru wkh xsvwuhdp jrrg lq wkh sduwldoo| lqwhjudwhg fdvh glhuv iurp wkh lqyhuvh ghpdqg lq wkh xqlqwhjudwhg fdvh1 Lq wkh xqlqwhjudwhg fdvh/ erwk grzqvwuhdp upv kdyh v|pphwulf ghpdqgv iru wkh xsvwuhdp jrrg1 Lq wkh sduwldoo| lqwhjudwhg fdvh/ f phdvxuhv rqo| vdohv wr wkh xqlqwhjudwhg grzqvwuhdp Ilup 51 Ixuwkhupruh/ wkh grzqvwuhdp sulfhv duh qrw v|pphwulf ehfdxvh wkh pdujlqdo frvw ri wkh lqwhjudwhg up lv }hur/ exw wkh xqlqwhjudwhg grzqvwuhdp Ilup 5 idfhv d pdujlqdo frvw ri Lq wkh vhfrqg vwdjh htxloleulxp/ % & % & qE w2 f ' k 2 +5:, 2E w E n w 2 2 Khqfh wkh lqyhuvh ghpdqg iru wkh xsvwuhdp jrrg lv % & n w Ef' 2 k 2 w f +5;, qE w2 2 2 69 Lq wkh uvw vwdjh ri wkh jdph/ Ilup 4*v ghflvlrq sureohp lv wr fkrrvh %c wdnlqj %2 dv jlyhq/ wr pd{lpl}h

Z ' R^ERcR2n% ' -ERcR2n% zkhuh R dqg R2 ghshqg rq wkurxjk wkh htxloleulxp ydoxhv jlyhq e| +59, dqg ' E% n %2 lv jlyhq e| +5;,1

Ohw -ERcR2 ' R^ERcR2 Wkh uvw0rughu frqglwlrq iru Ilup 4 lv # $ # $ _Z Y- YR Y- YR2 Y Y ' n n n % 'f +5<, _% YR Y YR2 Y Y% Y% Wkh vhfrqg eudfnhwhg whup lv wkh xvxdo uvw0rughu frqglwlrq iru wkh xs0 vwuhdp Qdvk0Frxuqrw frpshwlwlrq1 Wkh uvw whup lq wkh uvw eudfnhw/ Y- c YR lv }hur ehfdxvh wkh upv duh Qdvk0Ehuwudqg frpshwlwruv lq wkh vhfrqg vwdjh ri wkh jdph1 Wkxv wkh uvw0rughu frqglwlrq iru Ilup 4 lv # $ # $ _Z Y- YR2 Y Y ' n n % +63, _% YR2 Y Y% Y% # $ w Y ' R En n % 'f +64, E w2 Y% 2 Wkh uvw rughu frqglwlrq iru Ilup 5 lv wkh vdph dv lq wkh xqlqwhjudwhg fdvh/ # $ _Z2 Y ' n %2 'f +65, _%2 Y%2 Wkxv/ xvlqj +5;,/ wkh uhdfwlrq ixqfwlrqv lq wkh xsvwuhdp pdunhw duh

k qwR E % ' % +66, eE w eE w2 2 2 2 e dqg k % ' % +67, 2 eE w 2 2 Wrwdo vxsso| lq wkh xsvwuhdp pdunhw lv k qwR E % n % ' f ' +68, 2 E w SE w2 2 e 6: Ghpdqg iru wkh xsvwuhdp jrrg lv ^2 ' k qER2E wREc zklfk/ xvlqj +59, lv k qE w2 ^ ' 2 +69, 2 2E w 2E w2 2 e

Dorqj zlwk +59,/ +68, dqg +69, gh qh wkh vxsso| dqg ghpdqg ixqfwlrqv iru wkh xsvwuhdp jrrg1 Lw lv lqvwuxfwlyh wr frpsduh wkh htxloleulxp ri vxsso| dqg ghpdqg iru wkh xqlqwhjudwhg dqg wkh sduwldoo| lqwhjudwhg pdunhwv1 Wr gr wklv/ zh qrupdol}h wkh xqlqwhjudwhg vxsso| dqg ghpdqg htxdwlrqv iru wkh idfw wkdw wkh xqlqwhjudwhg zkrohvdoh pdunhw lv wzlfh dv odujh1 Iru wkh xqlqwhjudwhg fdvh/ zh kdyh k k qE w ' +6:, E w 2E w 2E w 2 2 2 Iru wkh yhuwlfdo lqwhjudwlrq fdvh/ zh kdyh +diwhu vrph uh0duudqjlqj k qwR E k qE w qw ' +6;, E w SE w2 2E w 2E w 2E w2 2 e 2 2 e

D frpsdulvrq ri htxdwlrqv +6:, dqg +6;, vkrzv wkdw wkhuh duh wzr hhfwv wkdw ghwhuplqh zkhwkhu wkh xsvwuhdp sulfh lv kljkhu ru orzhu diwhu yhuwlfdo lqwhjudwlrq1 Wkh uvw lv wkh vwudwhjlf hhfw dqg lv uh hfwhg lq wkh whup qwRE Wkh vwudwhjlf hhfw uhgxfhv wkh vxsso| ri wkh xsvwuhdp jrrg1 SE3 w2 e Ilup 4 zloo xqghu0lqyhvw lq fdsdflw| uhodwlyh wr wkh xqlqwhjudwhg fdvh lq wkh uvw vwdjh ri wkh jdph/ dqg e| dq dprxqw wkdw lqfuhdvhv zlwk w1 Wklv lv wkh udlvlqj ulydov* frvw hhfw1 Ilup 4 xqghulqyhvwv wr lqfuhdvh wkh sulfh ri wkh lqwhuphgldwh jrrg1 D kljkhu sulfh ri wkh lqwhuphgldwh jrrg fdxvhv wkh xqlqwhjudwhg grzqvwuhdp up wr lqfuhdvh lwv uhwdlo sulfh/ zklfk lqfuhdvhv wkh uhwdlo ghpdqg iru Ilup 4 dqg lqfuhdvhv Ilup 4*v sur w1 Wkh vhfrqg hhfw lv wkh vxevwlwxwlrq hhfw1 Yhuwlfdo lqwhjudwlrq uhgxfhv wkh grzqvwuhdp sulfh ri jrrg 4 +e| holplqdwlqj grxeoh pdujlqdol}dwlrq,1 Wklv/ lq wxuq/ uhgxfhv wkh ghpdqg iru jrrg 5 ehfdxvh wkh| duh vxevwlwxwhv dqg/ fhwhulv sdulexv/ uhgxfhv wkh htxloleulxp sulfh ri wkh xsvwuhdp jrrg11 Lq wklv fdvh/ wkh uhgxfwlrq lq ghpdqg iru wkh xsvwuhdp jrrg uhodwlyh wr wkh xqlqwhjudwhg fdvh lv jlyhq e| qw Wkxv/ zkhwkhu yhuwlfdo lqwhjudwlrq udlvhv ru orzhuv ulydov frvwv 2E3 w e ghshqgv rq wkh uhodwlyh pdjqlwxghv ri wkhvh wzr hhfwv1

6; Lq wkh yhuwlfdoo| lqwhjudwhg fdvh/ wkh htxloleulxp sulfh ri wkh xsvwuhdp jrrg lv % & k E n wE n w TU ' 2 25 +6<, q 25 E 35 5 zkhuh 5 w21 e Yhuwlfdo lqwhjudwlrq orzhuv wkh xsvwuhdp sulfh lq wklv h{dpsoh1 Lq rwkhu zrugv/ wkh vxevwlwxwlrq hhfw lv odujhu wkdq wkh vwudwhjlf hhfw1481Grzq0 vwuhdp sulfhv dovr idoo lq wkh yhuwlfdo lqwhjudwlrq fdvh uhodwlyh wr wkh xqlqwh0 judwhg fdvh1

48Ri frxuvh wkhvh uhvxowv fdq fkdqjh zlwk glhuhqw dvvxpswlrqv1 RVV +4<<3, dvvxph wkdw wkh yhuwlfdoo| lqwhjudwhg up fdq frpplw wr dq xsvwuhdp sulfh1 Xqghu wklv dvvxps0 wlrq/ zlwk sulfh frpshwlwlrq wkh xqlqwhjudwhg xsvwuhdp up zrxog pdnh doo ri wkh vdohv ri wkh xsvwuhdp jrrg/ exw wkh kljk sulfh zrxog qrqhwkhohvv ehqh w wkh lqwhjudwhg up e| udlvlqj lwv ulydo*v frvwv1 Lw lv qrw gl!fxow wr vkrz wkdw wkh yhuwlfdoo|0lqwhjudwhg up*v sur w0pd{lpl}lqj zkrohvdoh sulfh lv kljkhu wkdq wkh sulfh lq wkh xqlqwhjudwhg fdvh li lv qrw wrr odujh1

6< Dsshqgl{ E

Wkh yhuwlfdoo| lqwhjudwhg up*v pdunhw vkduh lv

^ Ek qER wR2 r ' ' c +73, ^ n ^2 Ek qER wR2nk qER2 wR zkhuh # $ k R ' n w c +74, 2 q

R2 ' +75, dqg 5 6 k n wE n ' 7 2 8 +76, E nq w2E n 2 n 4@ surylghg Vxevwlwxwlrq lqwr wkh h{suhvvlrq iru r jlyhv 5 6 r ' 7 8 +77, 2 E n wn k3q 2 knqw Dgg wr uhihuhqfhv=

Table 1: Summary Statistics of Market Concentration and Vertical Integration Variables for the Entire Panel of Data

Mean Standard Maximum 75th Median 25th Minimum Deviation Percentile Percentile Percent of Stations* that are 0.21 0.12 0.58 0.28 0.19 0.11 0.03 Company-operated

Percent of Stations that are 0.28 0.19 0.68 0.45 0.26 0.11 0.01 Lessee-Dealer

Percent of Stations that are 0.11 0.67 0.28 0.14 0.10 0.06 0.00 Dealer-Company-Supplied

Percent of Stations that are 0.18 0.13 0.46 0.27 0.17 0.06 0.02 Dealer-Jobber-Supplied

Percent of Stations that are 0.21 0.085 0.44 0.27 0.20 0.16 0.02 Independent Retailers

Number of 2.18 1.05 5.00 3.00 2.00 2.00 0.00 Vertically integrated Wholesalers 2.45 1.84 9.00 3.00 2.00 1.00 0.00 Number of Unintegrated Wholesalers 0.10 0.05 0.39 0.12 0.09 0.07 0.01 Average Downstream Market Share** for Integrated Suppliers 0.14 0.07 0.47 0.16 0.13 0.09 0.01 Maximum Downstream Market Share of Integrated Suppliers *Percent of Stations of each vertical contract type is defined as the number of stations with that contract type in the retail census divided by the total number of stations in the retail census, for each metropolitan area. **Downstream Market share for an integrated supplier is defined as the number of integrated stations in the retail census that the refiner owns, divided by the total number of stations in the retail census, for each metropolitan area.

Table 2: Correlation Coefficients for Upstream and Downstream Market Variables for Broad Panel Regression Percent Percent Percent Percent Number of Number of Average Maximum Company- Lessee- Dealer Independent Unintegrated Vertically Downstream Downstream Operated Dealer Owned - Retailers Refiners Integrated Market share for Market share of Branded Refiners Integrated Integrated Suppliers Suppliers Percent* Company- 1.00 -0.73 0.40 -0.25 0.33 -0.23 0.15 0.10 Operated

Percent Lessee- -0.73 1.00 -0.80 -0.17 -0.45 0.25 -0.21 0.02 Dealer

Percent Dealer 0.40 -0.80 1.00 0.03 0.28 -0.10 0.21 0.01 Owned – Branded

Percent Independent -0.25 -0.17 0.03 1.00 0.24 -0.08 -0.14 -0.28 Retailers

Number of 0.33 -0.45 0.28 0.24 1.00 -0.08 -0.09 -0.12 Unintegrated Refiners

Number of Vertically -0.23 0.25 -0.10 -0.08 -0.08 1.00 -0.22 0.13 Integrated Refiners

Average 0.15 -0.21 0.21 -0.14 -0.09 -0.22 1.00 0.83 Downstream Market Share for Integrated Suppliers

Maximum 0.10 0.02 0.01 -0.28 -0.12 0.13 0.83 1.00 Downstream Market Share of Integrated Suppliers

Table 3: Cell Means by Combinations of Market Structure Variables Dependent Variable: Quarterly average price of unbranded wholesale gasoline by rack less the spot price of crude oil

Few Integrated Many Integrated Suppliers Suppliers

Small Large Small Large Downstream Downstream Downstream Downstream Market Share Market Share Market Share Market Share

Few Few 19.75 22.49 20.70 22.22 Unintegrated Independent (1.02) (1.13) (1.40) (1.54) Suppliers Retailers

Many 18.33 22.29 17.31 18.65 Independent (2.36) (2.36) (0.96) (2.59) Retailers

Many Few 17.54 19.29 13.10 13.19 Unintegrated Independent (0.86) (0.88) (1.11) (1.09) Suppliers Retailers

Many 15.30 16.27 11.61 14.38 Independent (0.71) (0.66) (1.29) (1.40) Retailers

Standard errors in parentheses

Table 4: Broad Panel Regression Results Dependent Variable: Quarterly average unbranded wholesale price by metropolitan area, less the spot price of crude oil * Robust Standard Errors corrected for serial correlation and city-specific heteroskedasticity

Parameter Robust Standard P-Value Estimate Error

Intercept 22.56 1.118 0.000

Number of Vertically -1.588 0.244 0.000 Integrated Suppliers

Number of -1.006 0.152 0.000 Unintegrated Suppliers

Average 12.309 5.613 0.029 Downstream Market share for Integrated Suppliers

Market Share of -6.485 3.512 0.066 Independent Retailers

California Reformulated 5.782 1.055 0.000 Gasoline Requirement

Adjusted R-Square 0.287

Number of N=26 Observations T=18

*Newey-West standard errors are reported, correcting for serial correlation and heteroskedasticity. First order autocovariances for the wholesale margin time series were insignificant in each metropolitan area. Second order autocovariances were negative and significant, but small, in a few of the metropolitan area time series. Higher order autocovariances were all insignificant.

Table 5: Characteristics of Markets affected by Tosco-Unocal Merger Downstream Market Share is measured as percent of total stations in the metropolitan area

Tosco's Unocals's Tosco's Unocal Competed Distribution Pre-Merger Pre-Merger Post-Merger with Tosco at the Rack Downstream Downstream Downstream Distribution Rack Market Share Market Share Market Share San Jose 0.00 0.15 0.15 San Francisco 0.08 0.07 0.15 Fresno 0.01 0.11 0.12 Yes Los Angeles 0.02 0.16 0.18 Yes Sacramento 0.06 0.08 0.14 Yes San Diego 0.03 0.11 0.14 Yes Stockton 0.03 0.07 0.11 Yes Santa Barbara 0.01 0.19 0.20 Phoenix 0.29 0.02 0.31 Tucson 0.39 0.06 0.45 Reno 0.00 0.08 0.08 Yes LasVegas 0.06 0.05 0.11 Seattle 0.22 0.02 0.24 Portland 0.13 0.05 0.18

Table 6: Regression of Effects of Raising Rival's Costs Dependent Variable: Weekly average unbranded wholesale rack price for Tosco less the rack price in Phoenix.

Fixed Effects Random Effects

Parameter Parameter Estimate Estimate Intercept -0.512 -0.572 (0.484) (1.365) [0.384] [0.675]

Downstream 0.445 0.369 Market Contact (0.087) (0.080) with Independent [0.000] [0.000] Retailers

Number of -0.083 -0.403 Wholesale (0.278) (0.203) Suppliers [0.766] [0.048]

Market Share of 0.105 0.084 Independent (0.164) (0.059) Retailers [0.521] [0.156]

Autocorrelation 0.824 0.824 Coefficient

R-Square Within: 0.017 0.176 Between: 0.219 0.395

Number of N=12 Observations T=129 Standard Errors in parentheses. P-Values in brackets.

Table 7: Ordinary Least Squares Regression of Scatter Plot in Figure 6

Parameter Standard T-Statistic P-Value Estimate Deviation Intercept 0.3055 0.4551 0.6713 0.5172

Downstream Market 0.3634 0.1108 3.2786 0.0083 Contact with Independent Retailers

Adjusted R-Square = 0.4699

N = 12 F = 10.71 Probability > F = 0.008

Table 8: Fixed Effects Regression of Effects of Raising Rival's Costs Dependent Variable: Weekly average unbranded wholesale rack price for Tosco less the rack price in Phoenix.

Parameter Standard T-Statistic P-Value Estimate Deviation Intercept -0.671 0.498 -1.35 0.178

Downstream Market Share 0.198 0.039 5.13 0.000

Number of Wholesale -0.071 0.278 -0.26 0.796 Suppliers

Market Share of Independent 0.105 0.164 0.64 0.521 Retailers

Autocorrelation Coefficient 0.824

R-Square 0.018 Within: 0.011 Between:

Number of Observations N=12 T=129

Table 9: Regression of Effects of Raising Rival's Costs

Dependent Variable: Weekly average unbranded wholesale rack price for Tosco

Fixed Effects Random Effects

Parameter Parameter Estimate Estimate Intercept -0.041 2.475 (0.487) (1.565) [0.933] [0.114]

Tosco’s Unbranded 0.945 0.952 Wholesale Price in (0.012) (0.012) Phoenix [0.000] [0.000]

Downstream 0.372 0.296 Market Contact (0.090) (0.082) with Independent [0.000] [0.000] Retailers

Number of -0.010 -0.378 Wholesale (0.278) (0.203) Suppliers [0.971] [0.062]

Market Share of 0.177 0.094 Independent (0.166) (0.059) Retailers [0.288] [0.110]

Autocorrelation 0.827 0.827 Coefficient

R-Square Within: 0.798 0.943 Between: 0.172 0.397

Number of N=12 Observations T=129 Standard Errors in parentheses. P-Values in brackets.

U1 U2

D1 D2

Figure 1. Unintegrated Case U1 U2

D1 D2

Figure 2. Vertical Integration Case 2

x1(x2)ui Wholesale Supply, x Supply, Wholesale

x2(x1) x1(x2)vi

0 Wholesale Supply, x1 Figure 3. The Strategic Effect Unintegrated Supply

Integrated Supply Wholesale Price, W

wui wvi Unintegrated Demand

Integrated Demand

Xvi Xui Wholesale Supply, X

Figure 4. Strategic and Substitution Effects 1.8

1.6

1.4

1.2 theta = 0.6

1.0 theta = 0.4

Wholesale Price Wholesale 0.8 theta = 0.2

0.6

0.4

0.2

0.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Independent Retail Market Share

Figure 5. Wholesale Price Versus Independent Retail Market Share 4.00

3.50

3.00

2.50

2.00

1.50

1.00

0.50 Estimated Wholesale Price Increase Price Wholesale Estimated

0.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00

-0.50 Increase in Downstream Market Share Weighted by Percent of Stations Competing with an Independent Retailer -1.00

Figure 6. Estimated Wholesale Price Increase in Each Metropolitan Area Versus Increase in Retail Market Share